AFAIR, in Ty-5 amp a pair of GM-70 was driven by a pair of 6P3S (6L6G) through an interstage transformer. And GM-70 were driving GM-100 tubes.
Edit: here is a picture of GM-100. When I was a kid my uncle made a fish tank out of it.
Last edited:
I was wondering why not use a current source in place of the bulb to offset the bias current, and reduce the voltage to minimize the power dissipation.
About the time I was ready to hit the post quick reply button it dawned on me that to do so would shunt the voltage swing on the transformer dampening any output and killing it.
The current source would probably be a better choice for balancing the bias current, but no power dissipation improvement, and a more complex ckt.
I've got a pair of supposedly NOS 3-500Zs, do you think I could cram one in each FireFly chassis? What size lamps would I need?
About the time I was ready to hit the post quick reply button it dawned on me that to do so would shunt the voltage swing on the transformer dampening any output and killing it.
The current source would probably be a better choice for balancing the bias current, but no power dissipation improvement, and a more complex ckt.
I've got a pair of supposedly NOS 3-500Zs, do you think I could cram one in each FireFly chassis? What size lamps would I need?
Attachments
Also, 100mA into 2700 ohms only swings 270 volts peak positive at the plate, maybe more negative in A2 for a quite asymmetric current swing and less RMS plate swing.
Anyway it's been real...
"I was wondering why not use a current source in place of the bulb to offset the bias current, and reduce the voltage to minimize the power dissipation."
The official theoretical max efficiency specs for class A triode are (not counting filament power either):
Inductor loading: 50%
CCS loading: 25%
Resistor loading: approx. 12.5% (counting DC resistor loss, depends somewhat on the OT pri Z to triode Rp ratio) to 25% (not counting DC loss and using HV to get a CCS like effect) (but if you do include the DC resistor loss, using HV to get a CCS like effect can drop it as far as you can take it below 12.5%)
CCS loading and resistor loading require twice the B+ (or two B+s if Circlotron)
But due to the triode plate efficiency problem (minimum voltage the plate can be pulled down to under load current) these % are generally cut in half for real world class A1 triodes. Class A2 can do better than A1, getting closer to the theory max, but will run into g1 meltdown if pushed too far.
Realistically, only the inductor approach is sane. Or the schemes to use P-P stages to get SE sound.
The official theoretical max efficiency specs for class A triode are (not counting filament power either):
Inductor loading: 50%
CCS loading: 25%
Resistor loading: approx. 12.5% (counting DC resistor loss, depends somewhat on the OT pri Z to triode Rp ratio) to 25% (not counting DC loss and using HV to get a CCS like effect) (but if you do include the DC resistor loss, using HV to get a CCS like effect can drop it as far as you can take it below 12.5%)
CCS loading and resistor loading require twice the B+ (or two B+s if Circlotron)
But due to the triode plate efficiency problem (minimum voltage the plate can be pulled down to under load current) these % are generally cut in half for real world class A1 triodes. Class A2 can do better than A1, getting closer to the theory max, but will run into g1 meltdown if pushed too far.
Realistically, only the inductor approach is sane. Or the schemes to use P-P stages to get SE sound.
Last edited:
It would be significant improvement, if to use a current source. AFAIR, Smoking Amp, Michael Koster, Revintage and myself were experimenting with such topologies. With current source load you can get higher voltage swing and lower power dissipation than with resistive load. In my Alligator, as I mentioned before, I got even better efficiency, modulating a current source by a signal of opposite polarity. Similar topologies were discussed here by others.
What we did not discuss, a cathode load with a light bulb and a choke in series. I tried such an idea in order to use a choke with lower inductance than needed for lower frequencies, but gave it up.
By the way, here is a page with George's musings about CCS load: SE outputs
Sure! But you need some filament PS underneath. Do you think a SMPS would be Ok?
There was a circuit for a highly regarded amp floating around the internet about 10 years ago. It used a 211 in SE on one side of a P-P OPT and a second 211 configured as a "CCS" on the other side of the OPT.
It dawned on me at the time that simply wiring a second triode and biasing it up similar to the first except that it was undriven was an easy way to waste half of your output power in the Rp of the undriven tube. It is entirely possible to make a pentode based CCS with a high plate impedance to accomplish this. There were threads discusing this several years ago. I was thinking 845 for the output tube and 813 for the CCS. I got some cheap "845 SE OPTS" from the same guy that ripped everybody off on Ebay several years ago and I used them instead. They aren't half bad and the amp sounds very good.
Depending on how many extra tons of AC you add to your house, you could use the 3-500Z for the output tube, and a 4-400 for the CCS.
In my case I have about a dozen live 833A's. At 100 watts just for the filament I have given up on building a really big amp. We have been having record heat and my recently replaced central AC can't get the house below 81 degrees in the afternoon. I will stick to P-P sweep tubes. This seems like the most efficient way to make big power.
I must admit that Ciro's circuit is something new. I haven't seen the amp, and don't pay much attention to other peoples designs, but I give anyone credit that comes up with something unique. Look at it this way, assume that it is in reality a 10 watt GM70 SET amp. Can you buy one, or build one for $700? Yes, I agree that the Ebay text should be more accurate. Correct and accurate text that realistically explains the amp might even sell more of them.
I think what Anatolyi (Wavebourn) is mentioning are the modulated current source schemes. These can get 50% theoretical efficiency just like an inductance loaded SE. These approaches are of the type I refer to as P-P schemes to give SE sound. Numerous variations around. An interesting area. The simplest is to just build a P-P amp (class A mode) with a triode on one side and a pentode on the other (it's grid is driven still, not a CCS), balance the DC idle current and balance the effective gm's (adjustable grid drive for the pentode). The triode calls the shots as far as voltage goes (SE sound result), the pentode balances the DC in the OT (non-gapped P-P OT) and doubles the load Z seen by the triode (twice the power output and lower distortion). The usual P-P cancellation of even harmonics is largely avoided. Gets you at least 1/2 way to SE sound, likely more. Then there are other tricks that can get you most all the way to SE sound.
Last edited:
That's the main point! Correct and accurate description may sell much more than scandals on all audio related forums all over the world.
Buck Winding
Here is a way to use a low voltage CCS to balance the current in an OT. A buck winding has been placed in the available space remaining on the xfmr bobbin, in this case a SE OT. This way a 5 Volt power source can power the CCS. The CCS could also be a modulated current source to double output power.
This approach avoids the need for a doubled B+. And for a gapped OT, one can then remove the gap spacer (or just reduce it's thickness) to greatly improve the inductance for LF performance.
Here is a way to use a low voltage CCS to balance the current in an OT. A buck winding has been placed in the available space remaining on the xfmr bobbin, in this case a SE OT. This way a 5 Volt power source can power the CCS. The CCS could also be a modulated current source to double output power.
This approach avoids the need for a doubled B+. And for a gapped OT, one can then remove the gap spacer (or just reduce it's thickness) to greatly improve the inductance for LF performance.
Attachments
Last edited:
I tried something similar using the secondary with a low voltage CCS many years ago and a coupling cap to keep the DC out of the speaker voice coil, it did actually work, but not that well. The secondary got quite warm with the amp or so of dc I had to run through it to cancel the flux due to the quiescent current in the primary. This was a cheap PP transformer driven by a 45.. Also the supply voltage to the CCS has to be high enough to make sure that there is sufficient compliance voltage available at max output to prevent the CCS from cutting off the peak.. (that sounds really bad)
Maybe I missed a post somewhere, but I am wondering if we understand at all how this output stage operates?
The bulb definitely does consume power, however to calculate the actual output power you need to treat the bulb's shunt dynamic impedance as being in parallel with the 2.7K opt not as a separate entity.. In a sense it is part of the output AC load the GM70 sees, but to some degree the reverse might also be true. The thevenin equivalent impedance of the opt and bulb in parallel is 1.75K and hence the load the GM70 sees is 1.75K, not 2.7K and given the ratio of the impedances roughly 2/3 of the power should be delivered to the load, and 1/3 is wasted in the bulbs. Since the OPT is a toroid the DCR is probably quite low, as a swag probably well under 100 ohms.
Note that this is not really a cathode follower either - both the plate and cathode contribute to the audio output - I think the potential voltage swing across that transformer is far greater than we realize. Hint: Take a look at the audio voltage across the secondary of the OPT when the GM70 is driven towards cut off.. It's NOT zero. The voltage developed across that output transformer will be result of the sum of the currents flowing through it from both legs - Obviously because one leg is passive (undriven) the behavior in the current domain might be rather asymmetrical. It is not nearly as simple as it first appears.
I think we need to do a simulation and a better job of analysis than we have so far.. Perhaps it would be worthwhile for someone to build this circuit with parts on hand to see what it actually does.
I choose to believe Ciro's comments on output power based on his measurements until such a time as someone does a viable analysis or simulation and/or someone builds it.
Perhaps I have missed something fundamental that you all get and I am all wet..
Last edited:
For the Edcor GXSE15-16-10K shown, I was able to fit 24 turns of #20 wire onto the available space. I used Teflon insulated wire to take the heat. With 3600 turns on the primary and 24 turns on the buck winding, I need 7.5 amps thru the buck to neutralize 50 mA thru the primary. Using the secondary for 8 Ohm at 15 Watts then requires a peak voltage compliance of +/- 2.56 V for the CCS, so 7.5 VDC should work for the CCS. The buck winding has a resistance of 0.14 Ohms, so should dissipate 0.14R*7.5A*7.5A => 7.9 Watts. The CCS 7.5A*6.45V => 48.4 Watts. (6.45V = 7.5V - 0.14R*7.5V)
Since I only plan to use it for a 6BQ5 with 6 Watts output, the CCS only needs +/- 1.62 V compliance, reducing the CCS dissipation to around 7.5A*3.95V => 29.6 watts. (3.95V = 5V - 0.14R*7.5A)
Obviously all this pointing out how inefficient CCS balancing is. And resistive balancing would be hitting almost twice this loss per watt output.
Later I intend to try modulating the CCS to double power output and lower the loading (increase R) on the 6BQ5. Then measure the FFT to compare with the original CCS case. Also plan to try removing the gap spacer, which would no longer be necessary, and check the FFT again. If that all works well, then I will only be using P-P transformers for SE from then on.
-------------
On Ciro's circuit, the resistor leg does indeed absorb about 1/3 the power output. The tube side doesn't really matter what confijuration it's in (other than the grid drive required), its just a SE stage. Compute its off voltage to its on saturation voltage at 2X idle current and thats the peak to peak V swing. (obviously the grid has to go positive to get to 2X idle current) That V swing should jive with the 0 to 2x idle current swing across the equivalent loading (bulb plus OT in parallel) One of the confusing factors there was that the bulbs, as configured, will not support sufficient idle current to get the stated power output. Needs around 75 Watts of bulbs (linear R assumed, to draw 147 mA idle) instead of 50 Watts.
Since I only plan to use it for a 6BQ5 with 6 Watts output, the CCS only needs +/- 1.62 V compliance, reducing the CCS dissipation to around 7.5A*3.95V => 29.6 watts. (3.95V = 5V - 0.14R*7.5A)
Obviously all this pointing out how inefficient CCS balancing is. And resistive balancing would be hitting almost twice this loss per watt output.
Later I intend to try modulating the CCS to double power output and lower the loading (increase R) on the 6BQ5. Then measure the FFT to compare with the original CCS case. Also plan to try removing the gap spacer, which would no longer be necessary, and check the FFT again. If that all works well, then I will only be using P-P transformers for SE from then on.
-------------
On Ciro's circuit, the resistor leg does indeed absorb about 1/3 the power output. The tube side doesn't really matter what confijuration it's in (other than the grid drive required), its just a SE stage. Compute its off voltage to its on saturation voltage at 2X idle current and thats the peak to peak V swing. (obviously the grid has to go positive to get to 2X idle current) That V swing should jive with the 0 to 2x idle current swing across the equivalent loading (bulb plus OT in parallel) One of the confusing factors there was that the bulbs, as configured, will not support sufficient idle current to get the stated power output. Needs around 75 Watts of bulbs (linear R assumed, to draw 147 mA idle) instead of 50 Watts.
Last edited:
I'm getting some very interesting results from initial simulations of Ciro's circuit. Hopefully I will be able to derive some real understanding, but the initial results look like Ciro's reported output powers are achievable. Very, very large voltage swings are possible across that OPT. (Despite my earlier comments they appear rather symmetrical.) Based on a very shaky analysis and the resulting simulation significantly more than 700Vpp appears possible across the OPT - this may be optimistic but indicates output power in the range stated is possible.
Last edited:
You can get some mis-leading results easily if the tube is biased to draw the required 147 mA idle current (Uhh, actually a little more than 147mA idle is required since the bulbs will not drop to 0 mA at the positive signal voltage peak) because the OT will not be setting with 0 mA idle current (actually the OT won't be setting at 0V idle either if there is a little R in the OT ). So the OT will not be balanced, but the Sim will be showing the required +/- 394 V swing.
Lets see, with a 100 V still on the 50 W bulbs at signal peak, they would still be drawing 20 mA, so idle current needs to be increased to 147mA + 20ma => 167 mA to get the required +/- 147 mA swing. Well, this interacts back again in the calcs, add infinitum. Really needs more like 100 Watts of bulbs for pullup to the +500V. Make that line cord wattage from my earlier calc. more like 750 Watts draw to get 29 Watts audio per channel.
Lets see, with a 100 V still on the 50 W bulbs at signal peak, they would still be drawing 20 mA, so idle current needs to be increased to 147mA + 20ma => 167 mA to get the required +/- 147 mA swing. Well, this interacts back again in the calcs, add infinitum. Really needs more like 100 Watts of bulbs for pullup to the +500V. Make that line cord wattage from my earlier calc. more like 750 Watts draw to get 29 Watts audio per channel.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.