drdna wrote:

It will be a little project to modify the amplifier, but it may be educational. To optimize the Moondog circuit for the NOS 2A3 tubes, so they will give their best, I plan to buy a new OPT with primary impedance around 3.5-4K.

Romy the Cat wrote:

You use Edgar’s refrigerator subs that use own LF SS amps (that use EQ of I am not mistaken), so you do not use the full advantage of SET LF sound.

Romy the Cat wrote:

To construct 6C33C DSET juts for sake of upperbass driver doss not sound reasonable to me.

floobydust wrote:

As Romy pointed out, having high gain is not the same as power. If you're only barely cracking the volume control on your amplifier or preamplifier then it should be obvious that you have more gain than you need. If you're picking piece parts (either built, kit or implementing somebody else's design) then you need to manage the various gain of each component and determine how to best integrate them into a system. As far as the gain in the amplifier goes, it's a moot point IF you manage the overall system gain properly. I've done two SET designs over the past few years... and they are quite different on gain, yet have very similar specifications. There are both technical and financial reasons for each design but both have been optimized for performance, not gain. A brief word on power.... the 45 triode in a proper design will deliver a solid 2- to 2.5-watts and can be flat from 20Hz to 50KHz (or more) depending on the design, components and how it's actually built. A 2A3 triode will deliver up to 4.5 watts in the same application... please note that this is ONLY a 3dB change (double the power). As such, it's not really surprising that you (Adrian) realized little benefit from the 2A3 over the 45... you didn't gain very much (power). Having speakers which have higher sensitivity is a real requirement for using low power amplifiers. And even then, the last 1 to 2 octaves will require some additional help for anything beyond moderate listening levels. If you plan on using 2- to 5-watt SET amplifiers, you need to mate them with speakers of adequate sensitivity AND which will mate well with a low powered SET, of which most (if not all) have no negative feedback and have low damping factor. As for modifying your Moondog amplifiers, I would recommend against it. While you can effectively reduce the overall distortion and gain a bit of additional power, it's not worth gutting the amplifier and start swapping out expensive parts. If you're not happy with the amount of power now. You would likely be better off with a different (SET) amplifier which has enough power to meet your needs.

drdna wrote:

You are correct, though I believe the subwoofer covers only below 40 hz.

drdna wrote:

This is coming from the same person who has built a 6-channel DSET? Anyway, it was more a thought about optimizing the operation of each of these tubes and matching them to the drivers they would work.

Romy the Cat wrote:

drdna wrote: You are correct, though I believe the subwoofer covers only below 35 hz. So???

Romy the Cat wrote:

Did you mean to have one powerful driver and drive with it both 2A3 and 6C33C with DSET transformers? We use to play with this idea…

Paul S wrote:

There is also something to be said for continuity, so one does not wind up with a lot of "optimized" parts. I wonder whether having a different tube "signature" for each "channel"/driver would work out.

Paul S wrote:

If it was possible to do all bass SET, it would certainly be nice to do it, because the good SET bass is musically nonpareil.

drdna wrote:

So, my thought was that below 35 hz, there are many ways of thinking about amplification. Some might say SS amplification is helpful due to power demands of horn loaded 18" woofers. To use 6C33 might be a trade off but I do not know if overall it would be a benefit. From your experience, what is your thinking?

I would withhold my comments about it. However, you need to find out how much power and how much gain you need for you subs and your room. Borrow for someone any 6C33C and measure how far the 6C33C will be from class A2 at a level of max tolerable bass in your room.

drdna wrote:

It is a good idea to use the 6C33 for the woofer and subwoofer, but I worry that it would compromise the LF sound due to the demands of the subwoofer.

I do not see why subwoofer’s demands would compromise anything. We are taking about SET. It sucks constant current. If you are taking about loading demand then it is more complicated subject.  The keys in the decision to go for separate DSET for upperbass in not LF but HF. You need to know if the inductance of your upperbass returns anything into MF channel.  THAT was why I separated my upperbass and MF – I have written about it as it was the issue with my Fane 8M driver.  If you have the same problem than a simple solution would be to add just one more 2A3 with own OPT and driver from the same driver and driver from it the upperbass. In your case it would be even simpler. You have two stages on two halves with 6SN7 with too much gain. You might convert your amp into two stages 6SN7-2A3 and make each halve of 6SN7 to drive own 2A3: one upper bass and another MF. All that you need to worry if your PS would care another 60mA. In this setting you might load MF 2A3 with respect of best harmonics and the upperbass 2A3 with respect of most power (if it were necessary)

Anyhow, that all properly shell be in a different thread…

 The caT

floobydust wrote:

Plate supply = 350 volts<br>OPT Primary impedance - 3.5K Cathode bias resistor = 866 ohm Cathode bypass capacitor = 50uF Grid resistor to ground = 249 You need to select a new OPT.... I would suggest the H-20-3.5U.

Romy the Cat wrote:

The input will be absolutely unchanged from the “E” channel of my current 6-ch Milq.

http://www.goodsoundclub.com/PDF/6-Chennal_Melquiades_DSET_Amplifier.pdf

It will be the same filter, the same grid resistor bias, the same 6E6P at 200V but at 20mA inkstand of my current 35mA. Then it will be custom made by Bad Purvine, 75% nickel, limited 1.5H inductance, ~150Hz, 1:1, low turnes, low capacitance, inner-stage transformer.  The transformer will be driving two grids on two tube sockets. One will be 2A3, 6A3 tube sockets with switchable 2.5V/6.3ACV. Another is RE604, LK460, KL71403, PX25, YO186 sockets with 4V. The AC balance on output tubes filaments will be set by 100R 5W high precision attenuators with 25 turns, sitting on the filaments secondary. The YO186 or 2A3 will be used interchangeably, one per time. They will be powered with cathode bias from 400V and presumed 60-70V on cathode for YO186. It is not clear if 2A3 would allow to use the same cathode resistor, if not then it will be a switch to add one more resistor in parallel to set on 2A3 approximately 40-50mA. I would like intentionally to drive 2A3-type tine with lower current in order to drive their plate impedance higher and to make it closer to YO186/RE604 type of tubes as I will have just one OPT transformer that will be running both plates and it will be DSET-type, with limited DC current and limited inductance. I do not care about the current clipping on 2A3 tube as I will have a low of wasted gain with my only one Vitavox S2 driver as load, so it will be VERY far from running at full power.  The DSET output transformer is made by David Slagle, 80% nickel, limited 5H inductance, ~150Hz, 20:1, low turns, and low capacitance. 

So, as the result I shall be able to have a little blood to drop but I will be able to drive all tubes that I would like to drive with juts puling one tube out and plugging another tube in. I do not see that the “universally” of this design implies any compromised in ether2A3 or YO186 chin. The space in the amps do allows me to afford this luxury, so why do not take advantage of it?

Thanks, KM

It makes sense but I do not like to burn those tubes at max plate dissipation.  How about if I put one more RC chains in PS and drop anode voltages from 400V to 350V? Then if burn let say 60V in cathode then I arrive to a comfortable 290V on plate. If I run the tubes at 40mA then they both will be at 12W dissipation and it is what I would like them to be. I asked Dima to help me to calculate the circuit for this or for the similar to this operation. Can you check where your distortion numbers will bed? I think at 350V from PS I might be able to drive both tubes with just one switch of cathode resistor shunt.

About the 6A3. I did try it as I did not have any problem. Also, do not forget that I will be running it HF only and if a slightly higher 50Hz noise is there then it is not really auditable. Yes, the modulations might be there but is it the proof in a pudding?

floobydust wrote:

Romy,  As you're planning to run a 6K load, you can run less current without the transfer curve suffering as much. So, another shot at some numbers.... rather than use a RC setup to drop the voltage, Hammond (not my favorite) make an open frame choke rated at 9H and 40ma with a 700 ohm resistance (154G). Using this would drop close to 30 volts at 40ma of current plus whatever you lose in the primary of the OPT. From 400 volts down possibly to 365 volts.  Using a 1.5K cathode resistor and  a 370 volt supply, idle current would be ~39ma, plate dissipation at 12.25 watts and plate voltage at 309. Calculated distortion at 3.5-watts output is 4.6% 2nd harmonic and 0.2% at 3rd harmonic. The transfer curve looks decent and gets you very close to what you wanted.  Using the 45 with the 154G choke (LC for dropping voltage) and the 2K cathode resistor would yield the following: 32ma, 10-watts dissipation, 308 volts cathode-to-plate, 2.5-watts output with 5.3% 2nd harmonic and 0.2% 3rd harmonic. The transfer curve looks about the same as above.  In any case, you have a few options that all seem reasonable.  As for AC filaments on the 6A3.... not sure what your actual setup yields in measured output noise but would interesting to note which 6A3 tube you used and measured output noise. Using various 6B4G tubes it was unacceptable with AC. For output noise into an 8-ohm load I prefer a minimum of 80dB S/N referenced to 1-watt which equates to 283 microvolts of noise (maximum). You have a better chance at less noise from the 45 as it's only a single triode section where the latter 2A3 tubes are dual-sections strapped in parallel.

Yes, KM, how we are really close to the numbers were I would like to be. When I proposed to drive the plate voltage down I also proposed to use LC chain instead RC but Dima advised me do not do it. Listening his arguments I conclude that he is right. In this situation a choke or resistor would do juts a function of decupling of my DHT from PS. There is no filtration needed as I have enormous last cap in PS (I believe it 10K-20K in the end of LCRC). So, if I use let say a couple hundreds uF at my DHT tube’s anode then it will have around 1R impedance at the working frequency. The resole that I would put for in my C would be around 1000R, so the plate will be completely not informed about PS. In case I use a choke with DCR of 1000R then I have no advantages but I might pick some LF resonances due to the choke will be with very thin wire. So, it I have no advantages with choke but only possibilities for problems then why to use the choke? That were Dima’s arguments. I personally also would like do not use choke as it will be sitting inside of amp with all ugly magnetic radiations that I have none in my amp (all chokes and transformers in Milq are on PS side and there is no magnetic or high voltage AC in the control unit).

So, with RC I am not restricted with DCR of available or custom made chokes and I can driver plate voltage as low and I with as long as it benefits the best operation for both YO186 and 2A3. Who know I might find an operation mode for them to run on the same cathode resistor…

I think the next big task for me would be the identify the keystone of this operation the type and the volume of capacitors that will be sitting right besides the primary of the coupling transformer. I wish I can find 400V wet tantalum but I doubt….

floobydust wrote:

Question: the 0.75 ohm resistors in the filament lines on the 2A3. Is it your intent to drive them from a 2.5VAC transformer or a 5.0VAC transformer. Voltage drop considering the 2.5amp filament requirement would result in insufficient voltage to light it up.

floobydust wrote:

Comment: On the balancing scheme. Using a 100-ohm balance pot is quite common. Effectively this results in a 25-ohm resistance in series with the bypass cap. I also use a 100-ohm balance pot but pad it with a pair of 12-ohm resistors (end to wiper) which widens the adjustment range and drops the effective resistance down to ~4.8 ohms.

floobydust wrote:

Secondly, you might want to try a split balance approach. This has worked well for me with the 45 and 2A3 (dual-section types). I use a center-tapped filament transformer and connect the cathode resistor from the center-tap to ground. The balance pot has the padding resistors and the bypass cap only from the wiper to ground, i.e., you have a fixed DC balance and an adjustable AC balance.

Ok, thanks.

Ok, the Milq MF DSET is well laid out. I have ordered a few parts and it will be on it’s way. I’m still contemplating what kind 400V cap to put right before the DHT tube in B+. I have space and I might go do film cap but I am not sure. Still that cap is not a design decision but rather juts a decision.

Among the design decisions I have the only one dilemma to resolve. The two-stage Milq will have much more gain then I need for 109dB sensitive S2 diver. So, what I wonder is where to kill the gain. The right place to do it before the driver stage but I have a filter in there and this filter talks with positive bias of all 6 channels - it will be too messy to change it. Also if you look at the channel “E”:

http://www.goodsoundclub.com/PDF/6-Chennal_Melquiades_DSET_Amplifier.pdf

… then you see a filter resistor and bias resistor (was not possible to combined them as I did in other channels). So it have enough impedance on the driver’s grid and I do not want to add more as HF might suffer. What next? To deduce the gain of the driver stage is not good idea, probably it would be OK to burn the gain in coupling? Sounds uncomfortable and unpractical. Probably the ease way would be juts to put a 16R LPAD after the output transformer and deal with it, which bring me to a question: did anyone even ever seen not wire-wound LPAD but 16R stepped attenuators LPAD?  In past I had Taiwanese folks who built for me any crazy custom LPAD but they refused to humiliate themselves with low impedance and those idiotic low values of resistors. At that time I did not offer more money but now I would buy a pair of very high quality 16R, constant impedance, stepped attenuators.

Do anyone know what I might source them?

floobydust wrote:

As you pointed out, if the input circuitry and driver operating points are fixed, your only options are to lose it via coupling or the output stage. Maybe split them, i.e. use a 1:0.5 ratio on the inter-stage transformer and you could lose 6dB there. The rest you could lose at the output stage by using a L-pad to dial-in your preferred matching level and then replace it a couple fixed resistors.

Yep, the attenuation via L-pad and then with fixed resistors is fine but I do not feel that it is strategically kosher way to do the things.  Let pretend that I need to kill 9dB. I can do it after a transformer but it also would mean that I will send 9dB more voltage to grid of the output tube driving it closer to A2. If I have 1V at input then 30 times gain in the driver stage makes the AC remarkably close to the grid voltage at the output stage. So, I do not know it is it a good idea to approach the grid currents and then burn gain in the speakers. I will do the speaker attenuation to find out how much gain I need to burn (I anticipate that it will be 10-12dB) and then will see not can I lose 12dB in coupling. Sucks, too much complexity for suck a simple circuit. Sucks, too much complexity for suck a simple circuit.