Hello…
Driving a 300b always seems to be a point of discussion, but I can’t seem to find a mature schematic for a 300b driving 300b - any out there?
(I also note that some designers, like Thomas Mayer have also made 211 driving 211).
Ideally I’m looking for a two stage 300b pse/pp amp with interstage transformer and an input sensitivity around 2v or thereabouts.
Driving a 300b always seems to be a point of discussion, but I can’t seem to find a mature schematic for a 300b driving 300b - any out there?
(I also note that some designers, like Thomas Mayer have also made 211 driving 211).
Ideally I’m looking for a two stage 300b pse/pp amp with interstage transformer and an input sensitivity around 2v or thereabouts.
I use 2a3 or 6C4C outputs rather than 300b, so a bit easier to drive. My setup, which I'm very happy with, is a 10Y input stage in filament bias. To bring up the gain I use a 1:4 input step-up. A Hammond 1140-LN-C or 1140-LN-D does nicely, studio quality. A 10Y has a mu of 8 which is the minimum you need. You could also use a 112A, 2P29L or 4P1L, all in filament bias. These work with a 2V input. The 4P1L needs a current of at least 25mA. The others are less fussy. I use resistor loads.
I don't see the point of 300b driving 300b
1. It's expensive
2. It only has a mu of 4, which is too low
3. There are better sounding driver valves, especially a 10Y.
I don't see the point of 300b driving 300b
1. It's expensive
2. It only has a mu of 4, which is too low
3. There are better sounding driver valves, especially a 10Y.
Hi Andy - hope you are well.
Your post makes complete sense, but given this why are some people driving 300b with 300b, or 211 with 211?
The input sensitivity must be high, if the mu is only 4?
I like the look of Lynn Olsons Amity, with the 7119 which has a mu of 24 is apparently only just enough for this two stage design. Lynn is now using 6SN7 and has a preference to it over 7119 (I’m not saying it’s as good as 10Y though).
Your post makes complete sense, but given this why are some people driving 300b with 300b, or 211 with 211?
The input sensitivity must be high, if the mu is only 4?
I like the look of Lynn Olsons Amity, with the 7119 which has a mu of 24 is apparently only just enough for this two stage design. Lynn is now using 6SN7 and has a preference to it over 7119 (I’m not saying it’s as good as 10Y though).
Interesting @directdriver 300b double gain stage, makes more sense - I’d love 845, but I’m unwilling to work with such high voltages.
I guess then, then some of the single gain stage designs you see (e.g Vinyl Savour) must be low gain and not take the output 300b to full potential? Hence like Andy says might as well use 2a3?
I’m aiming for something around 20w - so need Pse/PP.
I guess then, then some of the single gain stage designs you see (e.g Vinyl Savour) must be low gain and not take the output 300b to full potential? Hence like Andy says might as well use 2a3?
I’m aiming for something around 20w - so need Pse/PP.
Marantz using a pair of 845s as rectifiers was an odd choice. I think some old WE amps did that.
Last edited:
Driving 300b with 300b, or 211 with 211 comes maybe from the Sakuma designs? But no real reason to duplicate those.
For a rough and ready calculation, just multiply the gain of each stage. I aim for something over 100, or over 120 is better.
My setup is SUT = 4, times 10Y = 7 (resistor load), times 4 for outputs = 116 total.
Just a 6SN7 driver gives 20 times 4 = 80. That's a bit low.
I like to use just 2 valve stages rather than 3. The SUT is quite transparent, more so than all the third valve stages I tried. A SUT allows you to make an all-DHT amp. That was the big step forward in my valve designs.
I don't use interstages - I like to cap couple with teflon FT-3 caps, which are very transparent. I put the money into the SUT on the input instead. Unfortunately Hammond prices more than doubled this year. The 1140-LN types are still affordable from KGA Enclosures.
For a rough and ready calculation, just multiply the gain of each stage. I aim for something over 100, or over 120 is better.
My setup is SUT = 4, times 10Y = 7 (resistor load), times 4 for outputs = 116 total.
Just a 6SN7 driver gives 20 times 4 = 80. That's a bit low.
I like to use just 2 valve stages rather than 3. The SUT is quite transparent, more so than all the third valve stages I tried. A SUT allows you to make an all-DHT amp. That was the big step forward in my valve designs.
I don't use interstages - I like to cap couple with teflon FT-3 caps, which are very transparent. I put the money into the SUT on the input instead. Unfortunately Hammond prices more than doubled this year. The 1140-LN types are still affordable from KGA Enclosures.
One thing to be a little careful with step-up transformers is they multiply the Miller capacitance of the input tubes. The impedance multiplication is the square of the turns ratio, so a 1:4 transformer multiplies the Miller capacitance (typically 60 to 80 pF) by 16 times. That can make a real challenge for the preamp, because it must drive the cable capacitance in addition to sixteen times Miller capacitance. The preamp load could be as high as 1000 pF. Many preamps, solid-state or tube, don't sound good with a load that heavy.
The current design of the Karna, also known as the Blackbird, uses 6SN7 -> triode-connected 6V6 -> 300B with custom interstage transformers from Cinemag. Transformer ratios are 1.2:1 step-down from 6SN7 to 6V6, and 1:1.2 step-up from 6V6 to 300B. The 6SN7 is 8-10 mA per section (gain of twenty), and the 6V6 is at 32 mA per tube, with a gain of about eight. The 1:1 input transformer is bypassed and out of circuit for XLR inputs (which are direct-coupled), and only used for RCA inputs.
The current design of the Karna, also known as the Blackbird, uses 6SN7 -> triode-connected 6V6 -> 300B with custom interstage transformers from Cinemag. Transformer ratios are 1.2:1 step-down from 6SN7 to 6V6, and 1:1.2 step-up from 6V6 to 300B. The 6SN7 is 8-10 mA per section (gain of twenty), and the 6V6 is at 32 mA per tube, with a gain of about eight. The 1:1 input transformer is bypassed and out of circuit for XLR inputs (which are direct-coupled), and only used for RCA inputs.
Last edited:
Thanks for that, Lynn. So in my setup the 10Y has a grid-plate capacitance of 7pf and a gain of 7 with resistor load. So if Miller C = C(1+ AV) where AV is gain of stage, then it looks like 7(1+7) = 56. That times 16 = 900. Is this correct?
What effect would that have on the HF of the stage?
What effect would that have on the HF of the stage?
20 Watts using two 300B tubes?
Then I recommend Push Pull 300B, if 300B is your choice of output tube.
20 Watts from two parallel 300B is more difficult, more expensive, and a real Hernia maker (your surgeon will love you).
I have built 300B SE, 300B PSE, and 300B PP.
Lynn Olson always has good ideas.
I am so glad to see him post here.
I worked with Lynn, Matt, Ken Boehlke, and others in the Spectrum analyzer division at Tektronix.
Guess Who?
Output power . . . do not forget the output transformer . . .
A nice push pull output transformer that has 0.5dB insertion loss will give the following result:
22.44 Watts from the PP 300B tubes into the primary = 20 Watts out from the secondary.
Then I recommend Push Pull 300B, if 300B is your choice of output tube.
20 Watts from two parallel 300B is more difficult, more expensive, and a real Hernia maker (your surgeon will love you).
I have built 300B SE, 300B PSE, and 300B PP.
Lynn Olson always has good ideas.
I am so glad to see him post here.
I worked with Lynn, Matt, Ken Boehlke, and others in the Spectrum analyzer division at Tektronix.
Guess Who?
Output power . . . do not forget the output transformer . . .
A nice push pull output transformer that has 0.5dB insertion loss will give the following result:
22.44 Watts from the PP 300B tubes into the primary = 20 Watts out from the secondary.
Last edited:
Hi Andy, 900 pF sounds about right. The preceding stage, or preamp, has to drive that plus whatever cabling is involved. Only a few pF if it is inside the chassis, but more like 200 to 300 pF if an interconnect is involved.
As for effect, capacitance does three things: a general HF rolloff formed by the RC of the source impedance and the total capacitive load, and an upper slew rate limitation when the preceding stage runs out of current to charge that capacitor. That's why it's desirable for drivers to be as linear as possible but also have a high current delivery capacity. And there's a third factor: a capacitive load creates an elliptical load-line, which creates much more distortion than a pure resistive load. (The lower part of the ellipse enters the low-current operating region of the tube, where it is least linear, and high-order distortion terms occur.)
One way to bypass the Miller C of a triode amplifier is use a pentode amplifier instead of a triode, but pentodes are known for their high-order distortion terms. A cascode circuit behaves similarly ... Miller C is nearly gone, but they have a pentode-like distortion spectra.
The alternative is brute force. More current, and as linear as possible. Resistive plate loads throw away about 1/3 of the gain, and load down the plate as well, so they are not well suited for driver applications. The alternatives are transistor current sources, inductor loading, or IT coupling. Transistor current sources are potentially colored-sounding, but more complex cascoded MOSFETs can get around this. If inductors are chosen, they must be audio grade and have minimized stray capacitance. IT's should be designed in consultation with the transformer designer.
If you are getting the idea that stray capacitance poses a lot of problems, you are right. It's worse with transistors, which have 10X as much capacitance, and it's nonlinear capacitance for good measure.
As for effect, capacitance does three things: a general HF rolloff formed by the RC of the source impedance and the total capacitive load, and an upper slew rate limitation when the preceding stage runs out of current to charge that capacitor. That's why it's desirable for drivers to be as linear as possible but also have a high current delivery capacity. And there's a third factor: a capacitive load creates an elliptical load-line, which creates much more distortion than a pure resistive load. (The lower part of the ellipse enters the low-current operating region of the tube, where it is least linear, and high-order distortion terms occur.)
One way to bypass the Miller C of a triode amplifier is use a pentode amplifier instead of a triode, but pentodes are known for their high-order distortion terms. A cascode circuit behaves similarly ... Miller C is nearly gone, but they have a pentode-like distortion spectra.
The alternative is brute force. More current, and as linear as possible. Resistive plate loads throw away about 1/3 of the gain, and load down the plate as well, so they are not well suited for driver applications. The alternatives are transistor current sources, inductor loading, or IT coupling. Transistor current sources are potentially colored-sounding, but more complex cascoded MOSFETs can get around this. If inductors are chosen, they must be audio grade and have minimized stray capacitance. IT's should be designed in consultation with the transformer designer.
If you are getting the idea that stray capacitance poses a lot of problems, you are right. It's worse with transistors, which have 10X as much capacitance, and it's nonlinear capacitance for good measure.
Last edited:
Thomas Meyer uses a 10Y/VT-25/801A DHT driver in his 300B designs, as well as his exotic DHT preamps. But 10Y/VT-25/801A's are not in current production ... they were discontinued way back in 1939. Maybe Elrog makes them in Germany, but I don't need to tell you Elrog is not cheap.
Thomas Meyer - 10Y
I used 45 drivers in the original Karna (Gary Pimm, is that you?), but they are plenty rare and expensive now, especially in matched pairs. It would be marvelous if Linlai or JJ made 10Y/VT-25/801A's or 45's, but I don't expect that to happen, considering how tiny the market is (guitar guys have no interest in these tubes, and they pay for the rest of the industry). Which is why the Karna/Blackbird has the tube lineup it has ... all tubes had to be in current production, along with reasonable NOS stocks for the tube rollers.
Which brings us to the Marantz T1 schematic. On close inspection, the input 300B's are at a relatively low voltage, with resistor loads, and a cap-coupled 1:2 interstage transformer. Not quite what you'd expect, and honestly, not the most linear operating region for the 300B's. But considering this design goes back to the mid-Nineties, Marantz didn't have much choice if they wanted an all-DHT amplifier.
Thomas Meyer - 10Y
I used 45 drivers in the original Karna (Gary Pimm, is that you?), but they are plenty rare and expensive now, especially in matched pairs. It would be marvelous if Linlai or JJ made 10Y/VT-25/801A's or 45's, but I don't expect that to happen, considering how tiny the market is (guitar guys have no interest in these tubes, and they pay for the rest of the industry). Which is why the Karna/Blackbird has the tube lineup it has ... all tubes had to be in current production, along with reasonable NOS stocks for the tube rollers.
Which brings us to the Marantz T1 schematic. On close inspection, the input 300B's are at a relatively low voltage, with resistor loads, and a cap-coupled 1:2 interstage transformer. Not quite what you'd expect, and honestly, not the most linear operating region for the 300B's. But considering this design goes back to the mid-Nineties, Marantz didn't have much choice if they wanted an all-DHT amplifier.
Last edited:
Lynn - some very useful posts, and many thanks. Your analysis of SUTs jogged me into replacing the shielded cable from DAC > amp with a short twisted pair. Indeed the sound improved, with brighter treble. I'm in your debt!
I was fortunate enough to buy up a bunch of 10Y when they were cheap, and still have some. It's my favourite small signal tube. No surprise there. I appreciate your need to use current production. Much better tubes could be used especially the DHTs, but also including some nice indirectly heated tubes like the 6AH4 which I always liked. At least there are reasonable numbers of them, though they're all in the USA.
I was fortunate enough to buy up a bunch of 10Y when they were cheap, and still have some. It's my favourite small signal tube. No surprise there. I appreciate your need to use current production. Much better tubes could be used especially the DHTs, but also including some nice indirectly heated tubes like the 6AH4 which I always liked. At least there are reasonable numbers of them, though they're all in the USA.
Very similar amp here, and my mods to the Blackbird design were largely based off measurement: frequency response, FFT, and static phase comparisons through the differential topology.
D/A is direct-to-pins transformer coupled with Jensen JT-11-DMPC; I think the nickel core is the way to go.
Allows for balanced to preamp, which is fully differential 12B4A (don't knock that triode, lots of current, gain not needed). Preamp output is LL1689AM/PP connected 9+9:2. I consider the Lundahl the weakest link in my system, but here it performs very well given the low Rp triode and easy to drive amplifier. 150 ohm output impedance is respectable for a tube preamp.
Main power amp:
Cinemag 2:1+1 input transformer, also only used with RCA inputs. Love that simple topology, go direct to grids when possible.
12SN7 fully differential with CCS->FT3->triode connected 46->Monolith bifilar IT-02->300B->Monolith B-8/6.6k.
My experience was the SN7 stage was extremely sensitive to load. No experiment with Lundahl or Onetics would provide proper phase balance at the plates; forget what you saw on the IT secondary; it was messing up the triode plates! For that reason I ditched the IT on the first stage and went with cap coupling; since the first stage is structured such that it is not heavily taxed even with clipped output at the speaker, I could get away with capacitors and not struggle with grid current and overload recovery associated with cap coupling.
I would be most interested in the Cinemag custom unit Lynn used, and how that performed. Subjective ears aside, how does that stage measure? Any knowledge of how the transformer is constructed?
Second experience was the thrilling discovery of 46/300B IT bifilar performance specific to the Blackbird push pull topology. Solved all the problems with phasing and frequency response until the 300B hits fairly heavy grid current. At that point, it's the 46 that struggles more so than the IT anyway. The 300V insulation rating is a restriction that needs to be accommodated with the bifilar design, but I was happy to constrain my build to gain the tight coupling between primary and secondary.
Stray capacitance, whether in the preamp circuit or within the transformer coupling, is no small feat to overcome. It is no argument that listening tests are always performed (who doesn't listen to their system), but also are measurements to find these shortcomings and maximize performance at each stage. One might easily enjoy the 'relaxed' sound of a compromised frequency response, not realizing your preamp is failing to properly drive its load. One could also prefer the 'musical accuracy' of having 2nd harmonic distortion in their system, believing they are hearing instruments as they were meant to be heard, not realizing the system is distorting and adding extra distortion not present in the original.
D/A is direct-to-pins transformer coupled with Jensen JT-11-DMPC; I think the nickel core is the way to go.
Allows for balanced to preamp, which is fully differential 12B4A (don't knock that triode, lots of current, gain not needed). Preamp output is LL1689AM/PP connected 9+9:2. I consider the Lundahl the weakest link in my system, but here it performs very well given the low Rp triode and easy to drive amplifier. 150 ohm output impedance is respectable for a tube preamp.
Main power amp:
Cinemag 2:1+1 input transformer, also only used with RCA inputs. Love that simple topology, go direct to grids when possible.
12SN7 fully differential with CCS->FT3->triode connected 46->Monolith bifilar IT-02->300B->Monolith B-8/6.6k.
My experience was the SN7 stage was extremely sensitive to load. No experiment with Lundahl or Onetics would provide proper phase balance at the plates; forget what you saw on the IT secondary; it was messing up the triode plates! For that reason I ditched the IT on the first stage and went with cap coupling; since the first stage is structured such that it is not heavily taxed even with clipped output at the speaker, I could get away with capacitors and not struggle with grid current and overload recovery associated with cap coupling.
I would be most interested in the Cinemag custom unit Lynn used, and how that performed. Subjective ears aside, how does that stage measure? Any knowledge of how the transformer is constructed?
Second experience was the thrilling discovery of 46/300B IT bifilar performance specific to the Blackbird push pull topology. Solved all the problems with phasing and frequency response until the 300B hits fairly heavy grid current. At that point, it's the 46 that struggles more so than the IT anyway. The 300V insulation rating is a restriction that needs to be accommodated with the bifilar design, but I was happy to constrain my build to gain the tight coupling between primary and secondary.
Stray capacitance, whether in the preamp circuit or within the transformer coupling, is no small feat to overcome. It is no argument that listening tests are always performed (who doesn't listen to their system), but also are measurements to find these shortcomings and maximize performance at each stage. One might easily enjoy the 'relaxed' sound of a compromised frequency response, not realizing your preamp is failing to properly drive its load. One could also prefer the 'musical accuracy' of having 2nd harmonic distortion in their system, believing they are hearing instruments as they were meant to be heard, not realizing the system is distorting and adding extra distortion not present in the original.
If you can't afford or put your hands on a 10Y, then a 46 is a good alternative. Not cheap these days, but better than 47 and 49 to my ears which come out cheaper. I tried a 46 in filament bias and liked it a lot - very dynamic sound. The 10Y was just more subtle. Filment bias has always been my preference, but the 46 was a challenge. It certainly needed big heatsinks on the Coleman filament regs, and the cathode resistors placed above the chassis.
Lynn,
I remember Gary Pimm too.
Matt and I built and wrote about "Not Your Father's Dyna" a single ended amp built on the Dyna ST70 chassis, for "Sound Practices".
Later, Matt and I did 600 measurements, collated them, and wrote "Paralleling Tubes Effects" for the last issue of "Glass Audio".
Hardly anybody noticed, so later I spoke about parallel tubes at one of the VSACs.
Then I did a presentation at VSAC 2008, with overheads of my PSE 300B amp and PP 300B amp, both with effective 1500 Ohm loads / plate, ECC99 driver to an interstage (SE to SE, and SE to PP interstages).
The second part of the presientation was the listening session, using the Left channel signal of Mono and Stereo CDs. The speakers were facing each other with one speaker connection out of phase, the volume of one amplifier was adjusted for a deep null. Then, the speaker was re-connected in phase, faced to the listeners, and a switch sent the music to SE, than PP, then SE, then PP.
There was one flaw in this method, I should have swapped the speakers to the other amplifiers, verified the null was still present, and then switched the signal to both the amp input and speaker output (using only one speaker for the listening test).
That would have eliminated the position change using two speakers, when I conducted the listening test.
Think about the one eared person, the person with a listener in front of him that masked one speaker's sound, etc.
I had to do the listening session, in order to realize I needed a speaker switch, not just an input switch.
Fun and Games!
I remember Gary Pimm too.
Matt and I built and wrote about "Not Your Father's Dyna" a single ended amp built on the Dyna ST70 chassis, for "Sound Practices".
Later, Matt and I did 600 measurements, collated them, and wrote "Paralleling Tubes Effects" for the last issue of "Glass Audio".
Hardly anybody noticed, so later I spoke about parallel tubes at one of the VSACs.
Then I did a presentation at VSAC 2008, with overheads of my PSE 300B amp and PP 300B amp, both with effective 1500 Ohm loads / plate, ECC99 driver to an interstage (SE to SE, and SE to PP interstages).
The second part of the presientation was the listening session, using the Left channel signal of Mono and Stereo CDs. The speakers were facing each other with one speaker connection out of phase, the volume of one amplifier was adjusted for a deep null. Then, the speaker was re-connected in phase, faced to the listeners, and a switch sent the music to SE, than PP, then SE, then PP.
There was one flaw in this method, I should have swapped the speakers to the other amplifiers, verified the null was still present, and then switched the signal to both the amp input and speaker output (using only one speaker for the listening test).
That would have eliminated the position change using two speakers, when I conducted the listening test.
Think about the one eared person, the person with a listener in front of him that masked one speaker's sound, etc.
I had to do the listening session, in order to realize I needed a speaker switch, not just an input switch.
Fun and Games!
I think our experiment with 100 Hy Cinemag inductors as 6SN7 plate loads was probably compromised by stray capacitance. The 2750 ohm DCR (yes, really) is probably a hint there are a LOT of turns in there.
How did you measure phase imbalance on the *SN7 plates? Direct measure with 100X scope probes, one probe for each scope channel, looking at 20 kHz?
How did you measure phase imbalance on the *SN7 plates? Direct measure with 100X scope probes, one probe for each scope channel, looking at 20 kHz?
Not that I know much about paralleling tubes, but it would seem like doubling Miller capacitance would be a big deal. And doubling the DC standing current going through the SE output transformer, which would require a larger air gap and a very large core to compensate. And the subtler issues of tube curves that are ever so slightly different between the paralleled pair. It seems like an awful lot of trouble for just a few watts.
The biggest problem in Class A PP is the initial phase split, which can be carried out at a low level, simplifying the problem.
The biggest problem in Class A PP is the initial phase split, which can be carried out at a low level, simplifying the problem.
I now do PP only with pentodes + cathode feedback and only use DHT's in SE or PSE. The PSE amplifier I see it in 2 ways:
1) one driver for the PSE output stage
2) individual driver for each output tube.
One driver for the the PSE output stage can be done with both transformer coupling and DC coupled cathode follower. DHT's are mostly low mu devices and once they are fairly selected and matched the difference in bias voltage can be addressed in a rather simple way: one fixed bias + individual small value filament bias resistor. This is very good solution especially if at same anode voltage and anode current the 2 tubes have near identical gm.
Running the Coleman's regulators for filament supply make life much easier and performance very high. For example, I run each tube of the 45 PSE at nominal 320V/30mA with 3.5K (7K per tube) for 5W @5% THD (2.5W @1% THD). Nominal bias voltage is -70V. Let's say, for example, one tube requires -69.5V and another -70.3V. I set the fixed bias at -69V, the total anode + filament current is 1.53A and so use 0.33R filament bias resistor on the tube that requires -69.5V and 0.51R on the other. Filament bias does not require bypass. So no capacitor arguments there....5-7W high quality resistors will do it. No large power resistors, heatsinks etc. are required like in 100% filament bias scenario.
Individual driver for each output tube would be like the Audio Note style amp with one DC coupled cathode follower for each power tube. It doubles up the coupling capacitors to the voltage amp stage but Miller capacitance seen by the voltage amp is very low anyway.
Pick your poison! 😀
1) one driver for the PSE output stage
2) individual driver for each output tube.
One driver for the the PSE output stage can be done with both transformer coupling and DC coupled cathode follower. DHT's are mostly low mu devices and once they are fairly selected and matched the difference in bias voltage can be addressed in a rather simple way: one fixed bias + individual small value filament bias resistor. This is very good solution especially if at same anode voltage and anode current the 2 tubes have near identical gm.
Running the Coleman's regulators for filament supply make life much easier and performance very high. For example, I run each tube of the 45 PSE at nominal 320V/30mA with 3.5K (7K per tube) for 5W @5% THD (2.5W @1% THD). Nominal bias voltage is -70V. Let's say, for example, one tube requires -69.5V and another -70.3V. I set the fixed bias at -69V, the total anode + filament current is 1.53A and so use 0.33R filament bias resistor on the tube that requires -69.5V and 0.51R on the other. Filament bias does not require bypass. So no capacitor arguments there....5-7W high quality resistors will do it. No large power resistors, heatsinks etc. are required like in 100% filament bias scenario.
Individual driver for each output tube would be like the Audio Note style amp with one DC coupled cathode follower for each power tube. It doubles up the coupling capacitors to the voltage amp stage but Miller capacitance seen by the voltage amp is very low anyway.
Pick your poison! 😀