Yup, tradeoffs. You're correct right there. And most of what you've pointed on both sides can be counter argumented on both sides.
Yes, with cathode bias your bias does shift with amplitude to some extent. And so it should with fixed bias, since your B+ will sag with higher amplitudes. This actually was a learning point for me!
When I designed my first class AB ultralinear PP amp, I used a separate transformer for bias, since my main PT did not have a bias winding. And that was kind of a mistake, since it made the bias supply too stiff! When the B+ sagged during louder passages, the bias didn't, so the amp goes into crossover distortion if driven to the edge. To be honest, this rarely happens, but it is a true phenomenon. Only when driven over clipping does this occur. But I still consider it a downfall of that design.
After that revelation I learned to use a bias winding tapped from the main secondary winding, so the sagging of the B+ is reflected to the bias voltage. And unsurprisingly the crossover distortion at clipping went away.
But this might be slightly counterintuitive. Nevertheless the moral of the lesson is to never regulate your bias supply unless you regulate your B+.
I've never really compared the same circuit design with both a cathode and fixed bias power stage, so I can't say much about the differences of those two. Both have their upsides.
Low frequency stuff is just a small part of a transformer's performance. Though not insignificant always... I've found reasonable PP trannies to work mostly unproblematically in the LF department. But high frequencies have shown to be a totally different universe. LF instabilities usually go away with considered coupling cap values, but HF instabilities sometimes take hours and hours with a scope and a signal generator to work out. My experience here is to never be happy with your measurements until you've gone up 'till a couple of megahertz... Output trannies sometimes may show resonances way up the spectrum.
But my point of "polishing a turd" had originally wasn't about a given transformer's LF response, even though sometimes that might have been just what the manufacturers were trying to correct. What I meant more, is that you have to design an amplifier that is already quite good without GNFB before implementing a global loop. Of course, this is all just belief based on a limited amount of experience, but it has gotten me somewhere from where I was maybe a decade ago (I used to be anti-NFB...).
Adolf,
Agreed,
Use a good topology and use good parts, before applying negative feedback.
As to distortion, many persons can live with some bass frequency harmonic distortion. Many loudspeakers have their own harmonic distortion.
Many recordings do not have very low frequencies in them.
Many composers write bass notes that are no more than 1 Octave below the frequency of the next note up in the music.
Many recordings are not of 32 foot organ pipes, nor Thunder Drums.
But . . .
Whenever the bass notes are saturated, then the Bass notes show up as upper and lower sidebands on each musical note, and on each musical notes harmonics.
A flute at 4kHz, and with 8kHz and 12kHz harmonics, will each have upper and lower sidebands on them (spaced away from the 4, 8, and 12kHz).
40Hz bass note that is causing saturation; 3960Hz, 4000Hz, 4040Hz; 7960Hz, 8000Hz, 8040Hz; 11960Hz, 12000Hz, 12040Hz.
Is that a flute I am hearing?
I am not sure.
How about the Diva singing a concert A (440Hz)? She sounds a little sick today?
No, just IMD sidebands.
Those are Intermodulation Distortions due to the core saturation.
All the above are just my opinion.
I am currently designing a 2 stage amplifier that uses what I call 'semi-global' negative feedback.
It takes the signal from the output tube plate / transformer plate connection. But it is not Schade feedback.
It is another technique that has often been used by others; output plate to input/driver tube cathode circuit.
I like the idea of not having to compensate for the output transformer's primary to secondary leakage inductance.
Agreed,
Use a good topology and use good parts, before applying negative feedback.
As to distortion, many persons can live with some bass frequency harmonic distortion. Many loudspeakers have their own harmonic distortion.
Many recordings do not have very low frequencies in them.
Many composers write bass notes that are no more than 1 Octave below the frequency of the next note up in the music.
Many recordings are not of 32 foot organ pipes, nor Thunder Drums.
But . . .
Whenever the bass notes are saturated, then the Bass notes show up as upper and lower sidebands on each musical note, and on each musical notes harmonics.
A flute at 4kHz, and with 8kHz and 12kHz harmonics, will each have upper and lower sidebands on them (spaced away from the 4, 8, and 12kHz).
40Hz bass note that is causing saturation; 3960Hz, 4000Hz, 4040Hz; 7960Hz, 8000Hz, 8040Hz; 11960Hz, 12000Hz, 12040Hz.
Is that a flute I am hearing?
I am not sure.
How about the Diva singing a concert A (440Hz)? She sounds a little sick today?
No, just IMD sidebands.
Those are Intermodulation Distortions due to the core saturation.
All the above are just my opinion.
I am currently designing a 2 stage amplifier that uses what I call 'semi-global' negative feedback.
It takes the signal from the output tube plate / transformer plate connection. But it is not Schade feedback.
It is another technique that has often been used by others; output plate to input/driver tube cathode circuit.
I like the idea of not having to compensate for the output transformer's primary to secondary leakage inductance.
Last edited:
Okay, I've been looking at power transformers, and I like the looks of this:
AS-2T350 - 200VA 350V Transformer - AnTek Products Corp
Looking at the loadline for a 6.6k transformer and a B+ of ~450v, current seems to peak at around 250ma. I would figure this would absolutely be enough transformer to run everything, and it's hardly more expensive than the 100va model, which might be pushing it.
I do plan to use SS rectification, so I think that I could just run the windings in parallel. If my math is correct, I think a 350v transformer should give around ~490 volts off of the rectifier. I plan on using a choke for the first filter after the cap, so I should be able to drop the voltage to around 450 volts under load.
Also, the Antek toroid doesn't have a bias winding. Should I use a small transformer attached to the filament to produce a bias voltage? I have some 10va doorbell transformers that have a 16volt secondary, which at 6.3 volts should give me about 45volts AC, or 66 volts DC, which seems like plenty for bias voltage, with a fair amount of adjustment. Does this seem unreasonable?
I guess I should try using PSUD2 to see what I can come up with.
Please let me know if anything seems unreasonable.
AS-2T350 - 200VA 350V Transformer - AnTek Products Corp
Looking at the loadline for a 6.6k transformer and a B+ of ~450v, current seems to peak at around 250ma. I would figure this would absolutely be enough transformer to run everything, and it's hardly more expensive than the 100va model, which might be pushing it.
I do plan to use SS rectification, so I think that I could just run the windings in parallel. If my math is correct, I think a 350v transformer should give around ~490 volts off of the rectifier. I plan on using a choke for the first filter after the cap, so I should be able to drop the voltage to around 450 volts under load.
Also, the Antek toroid doesn't have a bias winding. Should I use a small transformer attached to the filament to produce a bias voltage? I have some 10va doorbell transformers that have a 16volt secondary, which at 6.3 volts should give me about 45volts AC, or 66 volts DC, which seems like plenty for bias voltage, with a fair amount of adjustment. Does this seem unreasonable?
I guess I should try using PSUD2 to see what I can come up with.
Please let me know if anything seems unreasonable.
Last edited:
Hi 6A3sUMMER, which impedence of the speaker you take? Nominal, or you look at the impedance plot and see how it goes, then choose a value? Low frequencies have higher impedances, so flatter and not optimal loadlines.
1. An output transformer that is saturated can cause non-linearity all frequencies, not just bass frequencies.
Saturation of a push pull transformer is usually due to:
Either too much Un-balanced DC (more quiescent current from one plate than the other plate).
Or, when the very low frequency bass note is very large in amplitude.
The saturation depends on the amount of laminations versus the amount of the above causes.
2. A typical acoustic suspension 2-way 8 Ohm speaker impedance (Z) curve versus frequency might look like this:
At 20Hz Z = 6 Ohms, the DCR of the woofer voice coil.
Then Z increases as the frequency increases.
At 50Hz Z = 25 Ohms due to the woofer resonance of the woofer and air in the closed box.
Then Z decreases as the frequency increases.
At 400Hz Z = 6 Ohms, the DCR of the woofer voice coil.
Then Z increases as the frequency increases, due to the woofer voice coil inductance, and as the frequency gets nearer to the crossover frequency.
At 1500Hz Z = 16 Ohms, at the woofer to tweeter crossover frequency.
The Z decreases as the frequency increases above the crossover frequency.
At 4kHz Z = 8 Ohms.
The Z increases as the frequency increases all the way to 20kHz, due to the tweeter voice coil inductance.
3. A typical 2-Way 8 Ohm ported speaker impedance (Z) curve might look like this:
At 20Hz Z = 6 Ohms, the DCR of the woofer voice coil.
Then Z increases as the frequency increases.
At 30Hz, Z = 25 Ohms, due to the woofer and the port tuning.
Then Z reduces as the frequency increases.
At 50Hz Z = 6 Ohms due to the woofer DCR, and the port tuning that loads the woofer at that frequency.
Then Z increases as the frequency increases.
At 70 Hz Z = 25 Ohms due to the woofer and port tuning.
Then Z decreases as the frequency increases.
At 400Hz Z = 6 Ohms, the DCR of the woofer voice coil.
Then Z increases as the frequency increases, due to the woofer voice coil inductance, and as the frequency gets nearer to the crossover frequency.
At 1500Hz Z = 16 Ohms, at the woofer to tweeter crossover frequency.
The Z decreases as the frequency increases above the crossover frequency.
At 4kHz Z = 8 Ohms.
The Z increases as the frequency increases all the way to 20kHz, due to the tweeter voice coil inductance.
4. Notice that the typical impedance differences of acoustic suspension and ported enclosures is at the bass frequencies, not at the frequencies above about 400Hz.
Notice, that the impedance goes Below the nominal 8 Ohm impedance, as well as above the nominal 8 Ohm impedance.
And especially at the high impedance peaks, the impedance is inductive below the peak, and capacitive above the peak. These are RL and RC loads, respectively.
The loudspeaker is all kinds of RLC loading at various frequencies.
5. As to what tap to put the loudspeaker on, that is complex (no pun intended).
How much damping does the loudspeaker need?
What is the amplifier output impedance at a given tap? (hint, it is not the tap rating, unless the damping factor is very low, like = 1.0).
What is the distortion of the amplifier depending on the low impedances or high impedances of the load?
Saturation of a push pull transformer is usually due to:
Either too much Un-balanced DC (more quiescent current from one plate than the other plate).
Or, when the very low frequency bass note is very large in amplitude.
The saturation depends on the amount of laminations versus the amount of the above causes.
2. A typical acoustic suspension 2-way 8 Ohm speaker impedance (Z) curve versus frequency might look like this:
At 20Hz Z = 6 Ohms, the DCR of the woofer voice coil.
Then Z increases as the frequency increases.
At 50Hz Z = 25 Ohms due to the woofer resonance of the woofer and air in the closed box.
Then Z decreases as the frequency increases.
At 400Hz Z = 6 Ohms, the DCR of the woofer voice coil.
Then Z increases as the frequency increases, due to the woofer voice coil inductance, and as the frequency gets nearer to the crossover frequency.
At 1500Hz Z = 16 Ohms, at the woofer to tweeter crossover frequency.
The Z decreases as the frequency increases above the crossover frequency.
At 4kHz Z = 8 Ohms.
The Z increases as the frequency increases all the way to 20kHz, due to the tweeter voice coil inductance.
3. A typical 2-Way 8 Ohm ported speaker impedance (Z) curve might look like this:
At 20Hz Z = 6 Ohms, the DCR of the woofer voice coil.
Then Z increases as the frequency increases.
At 30Hz, Z = 25 Ohms, due to the woofer and the port tuning.
Then Z reduces as the frequency increases.
At 50Hz Z = 6 Ohms due to the woofer DCR, and the port tuning that loads the woofer at that frequency.
Then Z increases as the frequency increases.
At 70 Hz Z = 25 Ohms due to the woofer and port tuning.
Then Z decreases as the frequency increases.
At 400Hz Z = 6 Ohms, the DCR of the woofer voice coil.
Then Z increases as the frequency increases, due to the woofer voice coil inductance, and as the frequency gets nearer to the crossover frequency.
At 1500Hz Z = 16 Ohms, at the woofer to tweeter crossover frequency.
The Z decreases as the frequency increases above the crossover frequency.
At 4kHz Z = 8 Ohms.
The Z increases as the frequency increases all the way to 20kHz, due to the tweeter voice coil inductance.
4. Notice that the typical impedance differences of acoustic suspension and ported enclosures is at the bass frequencies, not at the frequencies above about 400Hz.
Notice, that the impedance goes Below the nominal 8 Ohm impedance, as well as above the nominal 8 Ohm impedance.
And especially at the high impedance peaks, the impedance is inductive below the peak, and capacitive above the peak. These are RL and RC loads, respectively.
The loudspeaker is all kinds of RLC loading at various frequencies.
5. As to what tap to put the loudspeaker on, that is complex (no pun intended).
How much damping does the loudspeaker need?
What is the amplifier output impedance at a given tap? (hint, it is not the tap rating, unless the damping factor is very low, like = 1.0).
What is the distortion of the amplifier depending on the low impedances or high impedances of the load?
Last edited:
To get a flat sound level response out of many speakers, they are dependent on equal voltages versus frequency.
But many speakers when driven by equal voltages versus frequency are not equal acoustic output.
I am not familiar with the Klipsch RF82.
I used to use a $50,000 Vector Network Analyzer I had access to at work to measure the impedance and phase angle versus frequency of many types of loudspeakers.
Then, I did measurements of the voltage and phase response versus frequency when those loudspeakers were driven from a non-feedback triode amplifier that had a damping factor of about 3 or 4.
Yes, there was some variation of voltage and phase versus frequency.
At work, I used a CD player and that same stereo amplifier and some of those speakers to play music at my desk.
There were several who commented on the good sound.
But many speakers when driven by equal voltages versus frequency are not equal acoustic output.
I am not familiar with the Klipsch RF82.
I used to use a $50,000 Vector Network Analyzer I had access to at work to measure the impedance and phase angle versus frequency of many types of loudspeakers.
Then, I did measurements of the voltage and phase response versus frequency when those loudspeakers were driven from a non-feedback triode amplifier that had a damping factor of about 3 or 4.
Yes, there was some variation of voltage and phase versus frequency.
At work, I used a CD player and that same stereo amplifier and some of those speakers to play music at my desk.
There were several who commented on the good sound.
Last edited:
Yes, 100VA would be pushing it somewhat. You should downrate transformers by a factor of 0.75 when rectifying, and you already have a minimum of ~20VA from the power tubes' heaters. Add around 100mA of idle current from the whole circuit, and you're at about 70VA. Would work, but not much headroom.
That 350VAC CT will end up in the ballpark of 460-480VDC when rectified and under load. And that's after a CLC filter with some but not much series resistance. Enough to land you some 30-40 watts with a UL connected EL34 power stage driving a 6k6 OPT.
And you should connect those two primaries in series to form a center tapped 350-0-350 secondary. That way you'll only need two diodes for a full wave bridge.
You could do that, or you could just use a separate bias transformer. Or then again consider whether using cathode biasing would be the wiser choice given you're trying to penny pinch?
As I hinted before, when using fixed bias the bias voltage should ride the power supply for sag, otherwise unwanted behaviour might occur when the amp is driven to heavy clipping. There are two ways to achieve this; either use a bias tap on the secondary, or use a cap coupled bias supply from a secondary without said tap. The latter cannot be loaded much, so I'd either go with a transformer with a tapped secondary, or just fudge it and use cathode biasing.
Just my two cents, but if you want to penny pinch, then use cathode biasing. You could then use that Antek toroid without a problem, and lose only maybe 5 watts of output power. With that PT and OPT you'll probably end up in the 30W ballpark with cathode biasing. Plus you don't have to build a very, very silent bias supply with all its associated components. Save moneys, lose not much. And owning and using the amplifier will be much easier. Especially if you'd like to experiment with "tube rolling".
If you ensist on using fixed biasing, I'd say go for a PT with a bias tap. Even if it costs a lot more, and it will. Probably double the Antek money. But the amp will behave nicer when deeply clipped, plus you save some chassis real estate. And you have a couple of holes less to drill. Going this route you might just hit maybe 37-40 watts at clipping with a fresh pair of JJ EL34's. And I'd call that a 35 watt amp...
If you want to go on the easy and cheap, there's more to gain on the cathode biasing side. But if you want the ultimate, there are some additional - though not proportional - gains on the fixed bias side.
Just my two cents...
imho, high bias shifts can be mitigated by the bypass cap paralleled to the cathode bias resistor...time constants can be made high enough....
music is such a high transient event....it is not as if bias will remain high for any appreciable length of time, you may not even catch it with your dmm..
music is such a high transient event....it is not as if bias will remain high for any appreciable length of time, you may not even catch it with your dmm..
Last edited:
Okay, in that case I may go with cathode bias bypassed with large caps. However, is there anything wrong with using one of the designs off the main winding like is shown on this page: The Valve Wizard
Thanks! If there is some good reason not to, I'll just go for cathode bias.
Thanks! If there is some good reason not to, I'll just go for cathode bias.
Actually, I use 60% for power factor (derating, effectively) for toroids. 100 VA = 60 watts DC load. Probably still “ok” as the amp won’t see full load all the time even when driven to clipping with music, and you’re not going to run sine waves for hours. But the 200VA are only a couple bucks more and not much physically larger so I use the 200’s a lot. The 200 and 400 VA Anteks are probably the biggest bang for the buck on the planet. They don’t offer enough voltage options on the 300’s.
I would use a 4 diode bridge with the windings in parallel. Yes, there are two diode drops but the transformer will sag much less under load using the whole winding on both half cycles. The leakage reactance is effectively dropped by 2X. You do end up with more voltage under load, which might put the amp solidly in the 40 watt class.
My experience with regulated g1 and g2 bias with unregulated B+ is that crossover distortion does not wildly go up when driven hard. In the big amp I’m working with, the bias does shift down during heavy clipping with a sine wave signal, but not much with music. It even takes several seconds to recover to quiescent after the signal is removed. The B+ rises back to 600V much faster (almost immediate). This points to a thermal memory effect in the tubes themselves causing the shift, not so much the drop in B+, which is about 12%. The 6550 curves are pretty darn flat from 500 to 600 volts. There is certainly more slope on the EL34, but it’s not like it’s the end of the world for a 10-15% drop in B+ If you’re far away from the knee.
Yep on real music with cathode bias you are OK if not driven into clipping. It does not make for a good THD graph. You can replace the cathode resistor with a zener (5W) and a smaller resistor in series with say 60-70% of the bias across the zener. Mr Summer calls this combination bias. This helps. With fixed bias you get more output power and you can run the tubes a bit cooler as there is no danger of crossover distortion. I tend to regulate the negative rail supplying the adjustment pots so it does not drop in sympathy with the HT/mains dips. You don't need much current. Fixed bias does require lower grid leak resistors than cathode bias (partially 6550's) so can have knock on implications on the driver.
Last edited:
1. If you use Regulated Fixed Bias, and do not Regulate the B+ too:
Then when the Mains Power Voltage rises, B+ rises . . .
That causes the output tube plate voltage to rise,
as well as the output tube screen voltage to rise.
And then Both the plate current and the screen current rise.
The tube quiescent dissipation increases.
I do not like that scenario.
I use un-regulated supplies. My mains voltage ranges from 117VAC to 123VAC.
A 300V quiescent B+ supply at 120VAC, will vary from 292.5VDC to 307.5VDC, that is a 15V range.
2. I am not the first one to call the zener plus resistor in the cathode circuit "combination bias". Others have called it that long before me.
But, just because I called it combination bias, does not mean I like it.
I will not use that kind of bias.
3. Over the years, I have used adjustable fixed bias, and self bias. I have even used battery bias (in a circuit that makes the battery live for years). Each has their tradeoffs.
I prefer Individual Self Bias for push pull outputs; all cases of parallel output tubes; and single ended single tube outputs.
Then when the Mains Power Voltage rises, B+ rises . . .
That causes the output tube plate voltage to rise,
as well as the output tube screen voltage to rise.
And then Both the plate current and the screen current rise.
The tube quiescent dissipation increases.
I do not like that scenario.
I use un-regulated supplies. My mains voltage ranges from 117VAC to 123VAC.
A 300V quiescent B+ supply at 120VAC, will vary from 292.5VDC to 307.5VDC, that is a 15V range.
2. I am not the first one to call the zener plus resistor in the cathode circuit "combination bias". Others have called it that long before me.
But, just because I called it combination bias, does not mean I like it.
I will not use that kind of bias.
3. Over the years, I have used adjustable fixed bias, and self bias. I have even used battery bias (in a circuit that makes the battery live for years). Each has their tradeoffs.
I prefer Individual Self Bias for push pull outputs; all cases of parallel output tubes; and single ended single tube outputs.
Last edited:
Just to echo other comments, I would absolutely avoid tapped secondaries. Go for either a single secondary in your most used impedance or Multiple secondaries.
I bought my Sowter UA21's configured as tapped secondaries which took the typical -3dB points of 5Hz to 50KHz for bandwidth (according to their data) and dropped the HF end down to a whisker under 30KHz for -3dB (I measured it at -3.3dB at 30KHz)
I bought my Sowter UA21's configured as tapped secondaries which took the typical -3dB points of 5Hz to 50KHz for bandwidth (according to their data) and dropped the HF end down to a whisker under 30KHz for -3dB (I measured it at -3.3dB at 30KHz)
I think point 1 is that regulated fixed bias does stop the bias voltage dropping if either the mains voltage drops or the HT drops and the winding is on the same transformer. The bias current is most sensitive to grid, screen and plate voltage in that order I think. I think you are saying that if the mains goes up then the increase in negative bias could protect you. However if the HT drops as a function of valve bias that will cause the g1 to get less negative and take more current I think. They work in opposite ways.
Last edited:
baudouin0,
If both the B+ and the Fixed Bias are not regulated . . .
Then for example, start with a 300V B+ and a -30V fixed bias with Mains Power at nominal voltage.
Let the mains voltage go up by 5%:
315V B+, and -31.5V bias
Let the mains voltage go down by 5%:
285V B+, and -28.5V bias.
More B+ makes more current, but more bias makes less current.
Less B+ makes less current, but less bias makes more current.
As you can see, it is somewhat self adjusting.
Does that seem reasonable?
If both the B+ and the Fixed Bias are not regulated . . .
Then for example, start with a 300V B+ and a -30V fixed bias with Mains Power at nominal voltage.
Let the mains voltage go up by 5%:
315V B+, and -31.5V bias
Let the mains voltage go down by 5%:
285V B+, and -28.5V bias.
More B+ makes more current, but more bias makes less current.
Less B+ makes less current, but less bias makes more current.
As you can see, it is somewhat self adjusting.
Does that seem reasonable?
Yep. I would think that the change to g1 is more significant. I think if you complete the proof with something like an EL34 at say 40ma and calculate the current delta biased for the negative bias change and the HT change on the screen and plate at the same voltage. Then we can calculate delta plate power for each.
Where is does not work is if the valves themselves are responsible for the drop in HT and the drop in negative bias. However it may be a much lesser effect.
Where is does not work is if the valves themselves are responsible for the drop in HT and the drop in negative bias. However it may be a much lesser effect.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.