I've experimented a lot with CCS-loaded pentodes with local feedback and have really been pleased with the results. The resulting distortion is low-order and very low level.
Attached is distortion spectrum from a CCS-loaded (10mA) EL84 with feedback set for gain of 10 and run at 10Vrms output. Ignore the spike between 8-9kHz. It's a measurement setup artifact.
An EL84 isn't as pretty as a DHT, but maybe it could be hidden behind a transformer so no one knows it's there. 😀
Attached is distortion spectrum from a CCS-loaded (10mA) EL84 with feedback set for gain of 10 and run at 10Vrms output. Ignore the spike between 8-9kHz. It's a measurement setup artifact.
An EL84 isn't as pretty as a DHT, but maybe it could be hidden behind a transformer so no one knows it's there. 😀
Attachments
I second the local feedback sound in pentodes. Downside is the drive challenge it creates on the preceding stage given the plate to grid impact on the input impedance
Ale
Ale
In the circuit above, I was using a p-channel fet follower to drive the feedback network to get a high input impedance and allow for all direct-coupling in the feedback loop. I also drove the top of the feedback network from the source of the CCS mosfet so that I wouldn't spoil the high open-loop gain of the CCS-loaded pentode.
The result was a circuit with high Zin, low Zout, and almost no distortion. Of course, it isn't going to appeal to everyone, especially not to those who are averse to mosfet followers in their tube amps or negative feedback loops.
Heres a sketch of a later experimental circuit tailored for an EL34:
Attachments
Sand......... hmmm, on a tube forum.....
My filament supply for my 300b outputs went funny today and is on the workbench, and I had to replace it with two bench DC supplies. At that moment, cursing away, I asked myself why I don't have a simple solid state amp in my possession. I actually don't, or even a simple tube amp apart from an old Leak Stereo 20 which I haven't used for years and no idea if it still works. I just have a highly optimised tube setup in at least 6 different chassis between the separate 46 and 300b stages and a PSU and filament supply for each.
No wonder I'm starting to like op-amps......
My filament supply for my 300b outputs went funny today and is on the workbench, and I had to replace it with two bench DC supplies. At that moment, cursing away, I asked myself why I don't have a simple solid state amp in my possession. I actually don't, or even a simple tube amp apart from an old Leak Stereo 20 which I haven't used for years and no idea if it still works. I just have a highly optimised tube setup in at least 6 different chassis between the separate 46 and 300b stages and a PSU and filament supply for each.
No wonder I'm starting to like op-amps......
Take it easy, use the simplest version .... with a little compromise, till A1 border. 😛
Attachments
Last edited:
Yes, the PMOS comes very handy here and is an extra device in the audio chain, however I’m a full supporter of the followers in any case
I did a similar experiment like what George came up with with the feedback divider network to the grid and the cathode driven by a PMOS. I got about 4W out of a 6e5p tetrode at less than 0.4%. Needed about 40Vpeak to drive so the driver (another 6e5p) did the job well. Overall harmonic profile was good and I like the sound. I used a 5k OPT. Arrangement is well suited for a PP amp as George proved lately, but I haven’t got that far on that path
Very interesting implementation. What’s you preference in driving the feedback network instead of the cathode?
Thanks
Ale
I was doing those experiments with the goal of making a driver that could drive a cathode follower output stage with low distortion. I eventually got up to over 320Vrms with only 0.06% distortion with an EL34.
One interesting thing I noted in my experiments was that the lower I made the screen voltage, the lower distortion got (until it was so low that I ran into the Vg=0 line). For the case that gave me 320Vrms at 0.06%, I was running ~1100V B+ with only 35V on the screen. It felt weird to do that.
Driving the cathode instead of the feedback network is functionally equivalent except that the screen to cathode voltage now varies. This may help overall linearity or it may hurt it (I think the latter is more likely, but I don't know). It would kind of be like UL, but could be pretty significant if you are running the screen as low as possible like I was.
The main reason I didn't try driving the cathode was that I didn't think of it at the time.
In my current DHT SE amp project I use a p-channel mosfet to drive the cathode of the pentode VA tube, the gate of which is driven by a feedback divider from output tube plate. This netted me 0.027% distortion at 1W, so I'm a fan of the approach now.
-Heath
Hi Heath
Excellent input and experience. Like learning from this!
I didn’t play much with input via cathode, will do if I get the chance in the future. I went down that path when attempting to avoid a negative supply and power everything other than filaments from HT.
You’ve got a clear point on the screen to cathode voltage variance. I didn’t think of that at the time.
Going back to the 300B, if anyone is keen to add an extra MOSFET (the PMOS) a good option as you suggest is to implement the driver with this topology. The negative supply is available, so just a few tweaks.
Now I’m tempted!
Cheers
Ale
Excellent input and experience. Like learning from this!
I didn’t play much with input via cathode, will do if I get the chance in the future. I went down that path when attempting to avoid a negative supply and power everything other than filaments from HT.
You’ve got a clear point on the screen to cathode voltage variance. I didn’t think of that at the time.
Going back to the 300B, if anyone is keen to add an extra MOSFET (the PMOS) a good option as you suggest is to implement the driver with this topology. The negative supply is available, so just a few tweaks.
Now I’m tempted!
Cheers
Ale
I like this very much.
I think to see an additional PS cap in the signal path for the AC input loop of the output stage (courtesy of the split rail input), but still, pull the input tube and the output goes max -ve bias. Could feed the input stage with a CCS fed shunt cap supply and really lock the OP stage bias voltage, would also take input tube variations (sample to sample) out of the equation. Hard to ignore these things.
I'm used to seeing the input stage with plate to signal common, still AC ground with the PS cap, but the split rail removes need to drop voltage from filament return to grid of OP tube.
Can I ask, did you build the circuit?.. my interest in this stems back to JC's 'upside down amp', was advised that the input transformer needed to be fairly specific (not necessarily off the shelf), but that is all beyond my understanding.
Of all the schems I see out here, I consistently finding myself appreciating yours, cheers for that.
HK
Last edited:
Ale,
In those those earlier experiments I was using power pentodes, since I was trying to make a very high swing driver and needed a high idle plate voltage. I did try power beam tubes and got significantly more distortion. This makes sense, power beam tubes tend to have kinky characteristics at low plate voltages and low currents. Those kinks may have gone away at low screen voltages? I don't know, but I got better results with true pentodes before I tried reducing screen voltage as much as possible.
Right now I'm playing with a small-signal beam tube (6BN11, dual version of 6EW6) and getting very good results. EL34 gave me an open-loop gain of 300 with CCS load. 6BN11 is giving me 2600. I've extended my feedback loop to come from the output tube plate (so that I can use a positive-bias, high impedance transmitting triode that would otherwise be unsuitable for an open-loop SE amp), and I'm getting very low distortion (0.027% at 1W).
I'm driving the cathode with the p-channel FET but I don't think there's much "UL effect" since screen is pretty high and swings are 1V at most here. In any case, overall distortion is 1/3 as much as if I directly connect the input tube cathode to the feedback network, which is the more standard configuration.
I'm loving how simple it is, but it definitely doesn't add color to the sound like the earlier experiments I ran that had ~0.5% distortion at 1W. It is just very much in control of the speaker and authoritative.
I'd like to see what this 6BN11 or a similar tube is capable of in the configuration of the experiments I ran. I bet that the high open-loop gain would reduce distortion even closer to nothing.
If you do choose to rig something up to play with, I'd be happy to offer details on what worked best for me, like CCS FET selection, etc.
Attachments
Sounds like heresy on a thread which I started to produce designs for an all-DHT 2 stage 300b SE, but I've been using indirectly heated drivers in triode for the last 2 weeks. First EL84 then EL33. Both surprisingly good sounding. And there are plenty of EL33 equivalents like KT61, 6P25, 6P26, EL41 etc. that I haven't even tried yet.
Once you have a mu of around 20 in the driver valve, you would then have enough gain if you used a 1:2 step-up input transformer. So my next question is please recommend a 1:2 step-up transformer for the input. Lundahl, Sowter, Jensen, Cinemag, Hammond Broadcast......? Plenty of models can do 1:2. So what is a truly inspirational SUT?
Give me some good suggestions, guys.
Once you have a mu of around 20 in the driver valve, you would then have enough gain if you used a 1:2 step-up input transformer. So my next question is please recommend a 1:2 step-up transformer for the input. Lundahl, Sowter, Jensen, Cinemag, Hammond Broadcast......? Plenty of models can do 1:2. So what is a truly inspirational SUT?
Give me some good suggestions, guys.
Last edited:
andyjevans,
A 1:2 voltage step up transformer does several things:
Its input impedance is 1/4 of the load impedance.
Its low frequency roll off is dependent on the signal source's output impedance versus the primary inductance.
Its high frequency roll off is dependent on both the distributed capacitance of the primary and the leakage inductance from primary to secondary; versus the output impedance of the signal source, and the impedance that the secondary drives, respectively.
There may be large phase shifts in the transformer (large group delay).
Given enough money, all the above problems can be overcome.
Get your wallet or credit card out.
A 1:2 voltage step up transformer does several things:
Its input impedance is 1/4 of the load impedance.
Its low frequency roll off is dependent on the signal source's output impedance versus the primary inductance.
Its high frequency roll off is dependent on both the distributed capacitance of the primary and the leakage inductance from primary to secondary; versus the output impedance of the signal source, and the impedance that the secondary drives, respectively.
There may be large phase shifts in the transformer (large group delay).
Given enough money, all the above problems can be overcome.
Get your wallet or credit card out.
What is:
a. Bias voltage of the output tube,
b. Measured in-circuit voltage gain of the driver, and
c. Max available input voltage to the grid of driver
d. Impedance of the voltage source to the grid of the input tube.
Try and avoid need to use transformer.
-
Otherwise, LL7903 as 1:2 - just make sure the source drive impedance is suitably low if you use RC circuit across the secondary. With open grid to signal common, drive impedance doesn't matter nearly as much.
HK
Last edited:
Yes, triode mode EL84 cannot fully drive 300B.
Before buying expensive input transformers, I would try pentode mode EL84.
Not bad, not bad ... checked !
BTW, LL7903 is about 250 euros/pair. (LL1544A is better choice for me)
For that money you can buy 5-6 pairs of great tubes like C3g, D3a ... etc.
and have the best driver for 300B for the rest of your life.
But...IMO
Before buying expensive input transformers, I would try pentode mode EL84.
Not bad, not bad ... checked !
BTW, LL7903 is about 250 euros/pair. (LL1544A is better choice for me)
For that money you can buy 5-6 pairs of great tubes like C3g, D3a ... etc.
and have the best driver for 300B for the rest of your life.
But...IMO
Last edited:
Hmmm... I'm still using my op-amp input with NE5534 and actually I have no need of an input transformer. The question is whether it would sound better than my op-amp solution. The answer may well be no. I was just putting it out there as a theoretical choice. Ale Moglia is using a x8 SUT, I think the LL7903, but with a driver mu of 20 the choice of SUT widens very considerably. But as you say, still expensive when I have an op-amp solution for almost nothing. Well, nothing in my case since I already have a few devices and PCBs. My current gain in this stage is x5.5 which is too much. I would think that x2 is enough. Just have to make sure the op-amp is stable at under x5 amplification. I don't know enough about op-amps unfortunately, but of course it's possible.
In my case, Wolfson WM8741 dac chip has LL1544A on output, then volume pot then 2SK170 buffer then C3g then 300B ...
I know what I'm talking about
I know what I'm talking about
Hello Andy! Rajko says it right. Remember - you wouldn't have entertained IDHT drivers a little while ago. So go the whole hog, try EL84 in pentode.
We will get you trying my shunt cascode driver eventually! That is where i have stayed, for 15 years.
We will get you trying my shunt cascode driver eventually! That is where i have stayed, for 15 years.
I must confess that I like the sound of an EL33. Like the EL84 but smoother. I have some E80L coming. These mu=20 medium power valves are a different animal from wimpy high mu pentodes in triode. I'm sure the 6V6 is good too - haven't yet tried it but have some.
I have a pair of LL1554, original not LL1554A. Shall I try and put these on the input? What do you suggest?
Rod - remind me where I find the schematic and explanation of your shunt cascade.
I have a pair of LL1554, original not LL1554A. Shall I try and put these on the input? What do you suggest?
Rod - remind me where I find the schematic and explanation of your shunt cascade.
Last edited: