Primary inductance and magnetic capability without core saturation limit the low frequency response at rated power. An SE transformer is at a disadvantage because the gap needed to avoid DC saturation reduces the inductance by up to 75%. For low frequency response, bigger is better. A feedback loop that includes the OPT can boost the low frequency response of some small transformers.
The high frequency response is limited by winding capacitance and leakage inductance. Both of these get worse as the transformer is made bigger. The leakage inductance will resonate with the winding capacitance to form a notch in the frequency response. This notch can be reduced by driving the transformer with a low impedance source. A feedback loop that does NOT include the transformer does this.
The upper limit for a good SE OPT is somewhere around 20 watts for a budget transformer. The 833A tube is a hard choice because of it's high plate resistance. I paid a good transformer winder a chunk of money to make me a 50 watt OPT for the 833A. Unfortunately the notch was at about 15 kHz. I never got back to trying it with local feedback but someday. Mags has proven that if you give a good transformer company enough cash and a forklift you can get good SE OPTs for the 833A that do 200 Watts.
The high frequency response is limited by winding capacitance and leakage inductance. Both of these get worse as the transformer is made bigger. The leakage inductance will resonate with the winding capacitance to form a notch in the frequency response. This notch can be reduced by driving the transformer with a low impedance source. A feedback loop that does NOT include the transformer does this.
The upper limit for a good SE OPT is somewhere around 20 watts for a budget transformer. The 833A tube is a hard choice because of it's high plate resistance. I paid a good transformer winder a chunk of money to make me a 50 watt OPT for the 833A. Unfortunately the notch was at about 15 kHz. I never got back to trying it with local feedback but someday. Mags has proven that if you give a good transformer company enough cash and a forklift you can get good SE OPTs for the 833A that do 200 Watts.
Tubelab,
Thank you for your explanation. It makes perfect sense.
The only thing I can think of is that the filament transformers are cheap and not stacked very well...I can see gaps where the "E" meet the "I" plates.(they have wedges of wood between the coils and core to keep the coils tight)...So maybe there are enough tiny gaps to keep them from saturating. Plus I'm using low power.
What can you expect...I got them from a surplus supply for $3.50 ea.
Thank you for your explanation. It makes perfect sense.
The only thing I can think of is that the filament transformers are cheap and not stacked very well...I can see gaps where the "E" meet the "I" plates.(they have wedges of wood between the coils and core to keep the coils tight)...So maybe there are enough tiny gaps to keep them from saturating. Plus I'm using low power.
What can you expect...I got them from a surplus supply for $3.50 ea.
Member
Joined 2009
Paid Member
At first this just sounds right, but actually it isn't. The Rp of a tube is related to it's Mu and its Gm. In fact you divide Mu by Gm to get Rp. Granted Gm goes up as the current is increased for a given tube, but in general if you want low Rp, you are choosing low gain tubes with high transconductance.
So a 45 has an Rp of 1650 ohms, and a triode wired 6L6GC has an Rp of 1700 ohms. Which eats the most current?
A pentode has a rather high Rp, but it can be lowered by application of local feedback to lower the gain.
So if you used half a PP transformer for an SE (with the other side (low voltage) DC biased to allow only a tunable 1-5mA imbalance) these problems would go away?
I guess the pros are a gain in simplicity over the PP - and tunability, but the con is running that bias from a tapped winding sitting at B+.
Has anyone tried this?
I guess running the OPT bias at B+ just requires a bit more isolation..
It wouldn't need a complex bias - if 12V would be enough (no idea of the DC resistance of the primary) a 50mA bias is only 0.6W - i.e. the numbers should be manageable. You could also add in some PSRR in there with a bit of thought - the bias side could also be the tunable B+ noise cancellation part.
I guess the pros are a gain in simplicity over the PP - and tunability, but the con is running that bias from a tapped winding sitting at B+.
Has anyone tried this?
I guess running the OPT bias at B+ just requires a bit more isolation..
It wouldn't need a complex bias - if 12V would be enough (no idea of the DC resistance of the primary) a 50mA bias is only 0.6W - i.e. the numbers should be manageable. You could also add in some PSRR in there with a bit of thought - the bias side could also be the tunable B+ noise cancellation part.
Last edited:
A better approach would be use a FET as an antitriode on the other side in a self splitter arrangement. This increases power delivery but the antitriode is entirely slaved to the SE triode preserving all of the triode character. This is because the FET is acting as a source follower and because of it much higher gm.
There were a number of interesting threads on this in the past.
Shoog
The low frequency behaviour is more about distortion than frequency response.
A 5K Se transformer with 35H will provide just below 4K primary impedance at 30Hz. Is there any difference in terms of distortion and frequency response if I compare a 50K plate impedance pentode and a 1.5K triode?
As often happens, the pentode might work actually better into a lower load than the triode in terms of distortion at medium-high Pout but this time it's about a reactive component that shouldn't be there.....
One can roughly see how good the combination is by evaluation of the ratio between the reactance XL=2*pi*f*L and REQ=Rp//Zp where Rp is the plate resistance and Zp the nominal primary impedance.
In the example above one can see that XL is the same in both cases and equal to about 6600 ohm at 30Hz. The Req is approx. 4500 ohm in the case of the pentode and approx. 1150 ohm in the case of the triode. So the ratio is 5.7 for the triode and barely 1.5 for the pentode and clearly the triode beats the pentode hands down because that ratio XL/REQ has to be as high as possible to get an undistorted signal (i.e. without a significant increase of distortion in comparison to 1KHz). In practice a ratio of 8-10 will be very good! Generally speaking, this is easily achievable (without relevant sacrifices in other areas) only with devices of moderate-to-low plate resistance.
However with pentode there is UL connection that reduces the plate resistance from 50K to 5-7K. Then, taking ad vantage of higher gain in comparison to triodes in the same power category, some local feedback as already mentioned.
As I said in the beginning FR is not really a big issue because also here the FR will be ruled by REQ. Although the triode is still better than the 50K pentode, the latter will still achieve -3dB at about 20Hz. That is not bad at all considering the absence of feedback and the fact that most speakers cannot do 20Hz....
A 5K Se transformer with 35H will provide just below 4K primary impedance at 30Hz. Is there any difference in terms of distortion and frequency response if I compare a 50K plate impedance pentode and a 1.5K triode?
As often happens, the pentode might work actually better into a lower load than the triode in terms of distortion at medium-high Pout but this time it's about a reactive component that shouldn't be there.....
One can roughly see how good the combination is by evaluation of the ratio between the reactance XL=2*pi*f*L and REQ=Rp//Zp where Rp is the plate resistance and Zp the nominal primary impedance.
In the example above one can see that XL is the same in both cases and equal to about 6600 ohm at 30Hz. The Req is approx. 4500 ohm in the case of the pentode and approx. 1150 ohm in the case of the triode. So the ratio is 5.7 for the triode and barely 1.5 for the pentode and clearly the triode beats the pentode hands down because that ratio XL/REQ has to be as high as possible to get an undistorted signal (i.e. without a significant increase of distortion in comparison to 1KHz). In practice a ratio of 8-10 will be very good! Generally speaking, this is easily achievable (without relevant sacrifices in other areas) only with devices of moderate-to-low plate resistance.
However with pentode there is UL connection that reduces the plate resistance from 50K to 5-7K. Then, taking ad vantage of higher gain in comparison to triodes in the same power category, some local feedback as already mentioned.
As I said in the beginning FR is not really a big issue because also here the FR will be ruled by REQ. Although the triode is still better than the 50K pentode, the latter will still achieve -3dB at about 20Hz. That is not bad at all considering the absence of feedback and the fact that most speakers cannot do 20Hz....
Globulator, have you ever heard about the Klimo Beltane? It was a 300B SE amp using a PP OT and one EL34 on the other side to balance the DC current. It was a good amp but surely not as good as a proper 300B SE with its properly gapped OT.
Last edited:
Shoog, the issue with that is that you no longer get the SE distortion characteristics. 2nd harmonic is reduced considerably so you lose the waterfall drop off in harmonics, and it will sound like a PP amp.
You can use a P-P OPT for SE, but it is not as simple as stuffing a DC voltage into the unused half. The DC voltage source is a low impedance, therefore it is a short for all the audio signal induced into that half, which ideally the same as the plate's signal. This in effect shorts out the AC signal on the plate leaving you with near zero power out.
What you need is a CCS load on the other half, set to the same current as is flowing through the output tube. The CCS must be rated for twice the B+ voltage. A tube or solid state CCS will work. I have tried this with a 10M90 and it does work, but I used cheap OPT's so I can't say how good it can be done.
Remember half of a P-P OPT is 1/4 of the OPT's rated impedance, so my 6600 ohm OPT is a 1650 ohm SE OPT.
I have heard of people running a high current DC through the secondary (speaker winding) to offset the tube current, but I have not tried this.
What you need is a CCS load on the other half, set to the same current as is flowing through the output tube. The CCS must be rated for twice the B+ voltage. A tube or solid state CCS will work. I have tried this with a 10M90 and it does work, but I used cheap OPT's so I can't say how good it can be done.
Remember half of a P-P OPT is 1/4 of the OPT's rated impedance, so my 6600 ohm OPT is a 1650 ohm SE OPT.
I have heard of people running a high current DC through the secondary (speaker winding) to offset the tube current, but I have not tried this.
This was discussed at some length and I never heard a conclusive decision on whether the distortion was preserved or not. Some said it was and some said it wasn't.
Shoog
Good point, well made.. it looks like the best way to use a PP transformer is to run a slightly unbalanced class A push-pull amp - who'd have thought it..
The 300B amp I mentioned in post #72......I breadboarded and then built an amp I called the 300Beast. It used guitar amp OPT's and other budget parts. The input stage was a 5751 LTP and the driver stage was an LTP made from a pair of SRPP's made with 6CG7's. SRPP's were in vogue at the time. It would remain in class A up to about 15 watts, verified with a scope across a 10 ohm cathode resistor in each 300B's hum pots ground lead. Power at clip in AB1 was 28 watts.
It had a bias pot for each 300B so DC balance could be set. Both LTP's had balance pots and there was an AC balance pot on the 300B's. The small tubes were all NOS, but the 300B's were used, and from different lots of first generation Sovtek's. Quite mismatched.
A good push pull amp with that many adjustments should be able to null the second harmonic down to better than -40db when driven with a clean 1KHz sine wave at a fairly low power (well into class A). It will degrade at frequency extremes, generally due to OPT imperfections.
The 300Beast could only get to about -25 db at 1KHz and dropped to about -10db at 10KHz. Oddly the second harmonic was better than -40 db at 20KHz. This is due to the loss through a $16 OPT at 40 KHz.
There should be no reason that amp sounded as good as it did, and it defied any and all attempts to improve it with better parts. It died of an exploded electrolytic several years ago, but I still have it stashed away, waiting for when I have time and a lab to dig deeper into it and analyze it stage by stage.
This one?
300 Beast | Tubelab
Those 300s cost a fortune! Probably why I like the GU50 - huge quality and power handling for peanuts!
Interesting the drivers were SRPP, I found my SE sounded rather dead and lifeless until I tried an SRPP driver arrangement - which made it really start to sing. Plus it meant one less big hot resistor to find
. I'm guessing it helps to have a big active swing in both directions for a driver.
Those 300B's were cheap. I got 5 used tubes from ESRC at a local (Florida) hamfest. All 5 for $100 USD. He guaranteed that at least 4 would work, all 5 did.
I had previously stated that I built this thing about 10 years ago. Thinking about it now, it was more like 15.
I agree that 300B's are too expensive now, but I still have some Chinese 300B's, 211's and 845's that I got back then when they were more reasonably priced.
I plan to keep the 300Beast intact until I can build something that sounds as good or better. My new lab won't be ready until sometime next year. I can't even guess when right now since the house it will be in is still under construction.
I have a pair of big Hammond 1628SEA OPT's that do not work well with the usual SE amp design. I still plan to try something unusual to get life out of them.......next year.
Member
Joined 2009
Paid Member
The SRPP is P-P as stated. It can be asymmetrical depending on the load impedance.
The 300B and many low impedance low Mu triodes require considerable drive VOLTAGE, about 100 to 150 Vp-p for the 300B depending on bias voltage. Some also have considerable Miller capacitance. You need some drive CURRENT to avoid slew rate issues charging and discharging Miller. As the grid approaches zero volts, but stays negative, some DHT's start to draw small amounts of grid current. Most DHT's and many other tubes respond well to A2 or AB2 operation. You need source tens (small DHT's) to hundreds (the 833A) of milliamps into the grid of the output tube from a positive voltage supply for A2 operation if this is desired. You only need to sink enough current to discharge Miller quickly enough to avoid slew rate limitations.
The SRPP does a decent job of providing the voltage and current capability needed here, but the distortion is higher than an ordinary common cathode stage, and direct coupling is problematic. Nevertheless, the 300Beast was a nice sounding very dynamic amp.
I have since learned to use mosfet drivers for most of my tube amp designs. A mosfet follower can be directly coupled to the output tube grid eliminating the coupling cap and its associated overload issues. It can source amps of grid current if needed, and offers lower distortion than most vacuum tube cathode followers.
I will do a comparison between the SRPP and a CCS/Mosfet follower (PowerDrive) in the 300Beast when I have time.
I'm a big fan of direct-coupled drivers.
The 6AS7G also tends to draw grid current while the grid is still negative, in our circuits about -15V WRT the cathode. We can drive the grid about +15V. Not considered a semiconductor for the task, as occasionally an arcing power tube can knock out the driver tube as well (although its pretty rare). If the semiconductor is adequately protected, it seems like it might be nice move.
The 6AS7G also tends to draw grid current while the grid is still negative, in our circuits about -15V WRT the cathode. We can drive the grid about +15V. Not considered a semiconductor for the task, as occasionally an arcing power tube can knock out the driver tube as well (although its pretty rare). If the semiconductor is adequately protected, it seems like it might be nice move.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.