Maybe some odd interaction with the grid bias supply? Does the same happen with those components out of the circuit? Wild guess....
anatech said:
Chris beat me to the same conclusion. If the C-L-C resonates at 120 Hz, the ripple will increase. Add more capacitance on the output side of your choke. You may wish to reduce the capacitance on the input side of the choke as well.
Please let us know the results of your tests.
Cris and Barry, I was begining to come to the same conclusion, that it could only be a resonance issue. Otherwise, how could the ripple be worse?
Well, I tested again by halving the input capacitance (two 47u in series) and doubling the capacitance on the output side of the choke. This did reduce the B+ ripple to about 1.3V and the output hum to 1.6mv. Both those values are still about double what I get with the original supply. So I think that bias supply has a significant effect on the performance. Barry, what's your take? I'm tempted to just go with it as is, and see how long the rectifier lives. With that 150 ohms added resistance, I may not be too far out of range.
rdf, I can't remove the bias supply and easily test the ps under close to operating conditions, otherwise I'd try it just to satisfy my curiosity.
Sheldon
Well, I tested again by halving the input capacitance (two 47u in series) and doubling the capacitance on the output side of the choke. This did reduce the B+ ripple to about 1.3V and the output hum to 1.6mv. Both those values are still about double what I get with the original supply. So I think that bias supply has a significant effect on the performance. Barry, what's your take? I'm tempted to just go with it as is, and see how long the rectifier lives. With that 150 ohms added resistance, I may not be too far out of range.
rdf, I can't remove the bias supply and easily test the ps under close to operating conditions, otherwise I'd try it just to satisfy my curiosity.
Sheldon
Sheldon,
Not to be a pest, but I would rea-a-a-ally like to know what exactly the waveform is of what you are encountering, since we are all wondering. A scope image possible? (Both sides of filter.)
While posting, I discovered in an old British GEC book on amplifiers that they connect a 10nF 1.5KV capacitor in serie with a 10K resistor across their power supply chokes in 100W+ amplifiers. This they say must not be omitted, but stated that it serves to take care of supply induced parasitics. I cannot imagine that this is your problem, but mention it just for completeness sake.
Not to be a pest, but I would rea-a-a-ally like to know what exactly the waveform is of what you are encountering, since we are all wondering. A scope image possible? (Both sides of filter.)
While posting, I discovered in an old British GEC book on amplifiers that they connect a 10nF 1.5KV capacitor in serie with a 10K resistor across their power supply chokes in 100W+ amplifiers. This they say must not be omitted, but stated that it serves to take care of supply induced parasitics. I cannot imagine that this is your problem, but mention it just for completeness sake.
Johan Potgieter said:
Hi John,
Pestiness is not a problem. The problem is, I don't have a scope. But, if I keep building this stuff, I'll have to get one - to satisfy myself in nothing else. In this case, the ps is supplying about 60 watts (360V@160mA). But the inductor is pretty small, so I don't know. I can try that out of curiosity.
Sheldon
BTW, anyone up for taking a shot a modeling the original ps and the choke ps? I don't think there is a way to model it is PSUD, so it would probably have to be a spice program.
Sheldon
Sheldon
I have done a quick Spice simulation using a sine wave ripple input of 15V, 100 Hz. Now we know the ripple is not sine wave, but it will show the trend of things for now.
I used several values for the choke and found no tendency to resonate. At the 350V (or such; I forgot) point I found a ripple of (obviously) 15V for 0Hy; 390mV for 1 Hy; 76mV for 5 Hy; and 38mV for 10 Hy. Others may find better values with a closer approximation. E.g. I did not include the power supply impedance (transformer winding plus rectifier resistance) for simplicity sake, but the tendency is clear.
I understand that you cannot get hold of an oscilloscope, but at this stage regrettably I cannot come to further immediate conclusions - perhaps others can see light, which I hope they will share with us. (I don't think though that we are looking at magnetic induction here; I have never found that to be much of a factor for a choke; rather more so for a power transformer mounted close to an output transformer.)
I used several values for the choke and found no tendency to resonate. At the 350V (or such; I forgot) point I found a ripple of (obviously) 15V for 0Hy; 390mV for 1 Hy; 76mV for 5 Hy; and 38mV for 10 Hy. Others may find better values with a closer approximation. E.g. I did not include the power supply impedance (transformer winding plus rectifier resistance) for simplicity sake, but the tendency is clear.
I understand that you cannot get hold of an oscilloscope, but at this stage regrettably I cannot come to further immediate conclusions - perhaps others can see light, which I hope they will share with us. (I don't think though that we are looking at magnetic induction here; I have never found that to be much of a factor for a choke; rather more so for a power transformer mounted close to an output transformer.)
Sheldon,
Just a little more info (too hot to sleep here): I inserted 120 ohm power supply resistance in serie with the 15V ripple source, and ran a sweep for resonance. That occurs for 1 Hy at 20Hz, and for 10 Hy at about 5 Hz with my simulation. But that peak is less than 6 dB. Ripple is reduced to about 25% of that previously mentioned with the addition of the power supply resistance. That will put fear of resonance to rest (and now hopefully me too).
Just a little more info (too hot to sleep here): I inserted 120 ohm power supply resistance in serie with the 15V ripple source, and ran a sweep for resonance. That occurs for 1 Hy at 20Hz, and for 10 Hy at about 5 Hz with my simulation. But that peak is less than 6 dB. Ripple is reduced to about 25% of that previously mentioned with the addition of the power supply resistance. That will put fear of resonance to rest (and now hopefully me too).
Thanks Johan,
Seems logical, but what do I know? Did you include the bias supply in your model? That seems to be critical, since the PSUD supply model without the bias supply gives about 10x (5v vs. 0.5v) the ripple that I actually see (this is ps w/o choke). The resistance on the ground lead is obviously a factor, but I'm still not sure how the main bias cap models. I'm guessing it should be the bias cap in series with 150 ohms, both in parallel with the input cap. The bias resistors would also show up as 150 ohm in series, before the input cap. Am I close here?
BTW, the amp works fine as is, notwithstanding the possible wear and tear on the rectifier. Mainly, I'd just like to understand what is going on.
Sheldon
Seems logical, but what do I know? Did you include the bias supply in your model? That seems to be critical, since the PSUD supply model without the bias supply gives about 10x (5v vs. 0.5v) the ripple that I actually see (this is ps w/o choke). The resistance on the ground lead is obviously a factor, but I'm still not sure how the main bias cap models. I'm guessing it should be the bias cap in series with 150 ohms, both in parallel with the input cap. The bias resistors would also show up as 150 ohm in series, before the input cap. Am I close here?
BTW, the amp works fine as is, notwithstanding the possible wear and tear on the rectifier. Mainly, I'd just like to understand what is going on.
Sheldon
I modelled the circuit that you gave in your post #13. As a load for the 350V I placed a 3K resistor across the 100uF cap; this draws 117mA. Slightly more than your amplifier would, but it will not make much difference to what we are looking for. As a load for the 255V I put 35K; just the one 255V output used. The only components I did not include was the divider for the heaters, which would not change the picture. To represent the transformer + rectifier I placed a 15V generator in serie with (later) a 120 ohm resistor, as mentioned.
To my reckoning this should give results close to what is expected. My first run was with a 100 Hz signal only - OK, you guys have 60 Hz so that should have been 120 Hz, but again the difference would just be one of scale. The second run was done with a generator with a frequency range of 1 Hz - 200 Hz, to see where any resonance was.
I hate loose ends! So given time (sometime the next week), I daresay I am going to rig this up and have a look myself with a scope so we can put this to rest. This is a simple thing which should not be a puzzle. I should have components here somewhere.
To my reckoning this should give results close to what is expected. My first run was with a 100 Hz signal only - OK, you guys have 60 Hz so that should have been 120 Hz, but again the difference would just be one of scale. The second run was done with a generator with a frequency range of 1 Hz - 200 Hz, to see where any resonance was.
I hate loose ends! So given time (sometime the next week), I daresay I am going to rig this up and have a look myself with a scope so we can put this to rest. This is a simple thing which should not be a puzzle. I should have components here somewhere.
Hi Johan,
Simple things are always a puzzle. Seldom do we get stuck on complicated things! The impedance of the transformer may prove to be important.
-Chris
Simple things are always a puzzle. Seldom do we get stuck on complicated things! The impedance of the transformer may prove to be important.
-Chris
Johan Potgieter said:
Wow Johan, I didn't mean to make you work so hard. I trust you are doing it not to fix the problem, as it really isn't the practical issue that made me start the thread. If you are, I'll feel bad that I got you into this. I was just curious as to the unexpected result. But,,,,, if you are willing to try it anyway, take out the choke and extra cap for a test too. I'm really curious as to why the ripple with the non-choke supply is as low as it is.
Maybe that design is more clever than it looks.
Sheldon
OK, let us get this done.
Circuit used is as per post #13, using a transformer winding with dc resistance = 87 ohm and a further 100 ohm in serie. I used a Si-bridge rectifier; this arrangement should not be very different from a GZ34 without extra resistor. Winding voltage was 312VAC.
Without choke at the +350V point (and with both 47uF and 100uF in parallel), ripple on scope was 5.4Vpp. At the bias input it was 16Vpp. This conforms appr. to the division of capacities over the bridge output. DMM readings (0.1uF in serie) were 1.66V and 4.7V respectively. As the waveforms were the usual triangle form after a rectifier, this also seems in order (rms not quite applicable).
[Interrupt: I am frowning; why is there not a C over the 2x75 ohm bias resistors, or did I miss that - if not, put a 220uF 25V one there.]
With only 100uF capacitor as asked in previous post, the values were 8Vpp and 15Vpp (bias side) respectively, again with appropriate DMM readings.
Inserting a choke (unknown; I guess about 7 Hy) ripple before the choke was 16Vpp and after (across the 100uF) 0.5Vpp. DMM readings again conformed (the ripple after the choke was a good sine wave). BTW, my 350V load drew about 100 mA.
Perhaps academic only, the ripple after the 3 x 3K3 (over the 47uF) was 2 mVpp on the scope and too small to measure on the DMM.
Now firstly (it did not register when I looked only at figures before) one would expect the ripple before the choke to be higher than without the choke because the C is lower; the choke impedance isolates the rest of the Cs. The value after the choke is of course quite lower with than without.
A last point: Our environments would probably differ, Sheldon, but I discovered that without proper earth on the power supply the DMM would read higher, even on the 1V+ readings, and significantly on lower scales, when I had the "live" probe in my hand. The serie 0.1uF (between the probe and thecircuit) has a finite impedance and hand capacitance and induction from outside play a role.
It was a nice little exercise to get basics dusted off again and I hope of value - no problem on this side. If not said, earthing the choke, or its orientation, had no effect in my set-up. Now, as Anatech so truthfully stated (post #31), for me back to the relatively "simple" exercise of sorting out my 130W distributed load output transformer design (QUAD type circuit) for class AB operation (a little more difficult than class A).
Circuit used is as per post #13, using a transformer winding with dc resistance = 87 ohm and a further 100 ohm in serie. I used a Si-bridge rectifier; this arrangement should not be very different from a GZ34 without extra resistor. Winding voltage was 312VAC.
Without choke at the +350V point (and with both 47uF and 100uF in parallel), ripple on scope was 5.4Vpp. At the bias input it was 16Vpp. This conforms appr. to the division of capacities over the bridge output. DMM readings (0.1uF in serie) were 1.66V and 4.7V respectively. As the waveforms were the usual triangle form after a rectifier, this also seems in order (rms not quite applicable).
[Interrupt: I am frowning; why is there not a C over the 2x75 ohm bias resistors, or did I miss that - if not, put a 220uF 25V one there.]
With only 100uF capacitor as asked in previous post, the values were 8Vpp and 15Vpp (bias side) respectively, again with appropriate DMM readings.
Inserting a choke (unknown; I guess about 7 Hy) ripple before the choke was 16Vpp and after (across the 100uF) 0.5Vpp. DMM readings again conformed (the ripple after the choke was a good sine wave). BTW, my 350V load drew about 100 mA.
Perhaps academic only, the ripple after the 3 x 3K3 (over the 47uF) was 2 mVpp on the scope and too small to measure on the DMM.
Now firstly (it did not register when I looked only at figures before) one would expect the ripple before the choke to be higher than without the choke because the C is lower; the choke impedance isolates the rest of the Cs. The value after the choke is of course quite lower with than without.
A last point: Our environments would probably differ, Sheldon, but I discovered that without proper earth on the power supply the DMM would read higher, even on the 1V+ readings, and significantly on lower scales, when I had the "live" probe in my hand. The serie 0.1uF (between the probe and thecircuit) has a finite impedance and hand capacitance and induction from outside play a role.
It was a nice little exercise to get basics dusted off again and I hope of value - no problem on this side. If not said, earthing the choke, or its orientation, had no effect in my set-up. Now, as Anatech so truthfully stated (post #31), for me back to the relatively "simple" exercise of sorting out my 130W distributed load output transformer design (QUAD type circuit) for class AB operation (a little more difficult than class A).
Yah, sorry folks, I just checked. I did not put in the 47uF directly over the rectifier output.😱
The readings then change significantly to 3.5Vpp at the +350V point (without choke) and 7Vpp for the bias, and accordingly for everything else, but the ratios with and without choke remains. Arguments remain the same, though. Apologies.
The readings then change significantly to 3.5Vpp at the +350V point (without choke) and 7Vpp for the bias, and accordingly for everything else, but the ratios with and without choke remains. Arguments remain the same, though. Apologies.
Johan Potgieter said:
Yes, that was the traditional method of dealing with the -Ldi/dt provoked by the rectifiers switching off as the mains passes through 0V in a choke input supply.
Johan Potgieter said:
Thank you for your effort and patience Johan. If I understand correctly, I think this all makes sense. Attached are my measurements made with DMM and 0.1uf cap in series. The AC readings indicated as a range were not stable (interaction with probe capacitance etc., as you mentioned?). The other values were stable. It seems we are within experimental error here.
At this point, the only thing I remain unsure of is how to adjust the rectifier cap input ratings based on actual circuit parameters, especially with that bias supply hanging out there. I suspect I'm not as far off as the first look might suggest, but haven't found any calculation method.
Again thanks,
Sheldon
Attachments
Regarding the voltage on the input capacitor (and actually all capacitors), one plays safe. The highest value that it can encounter will be the peak value of 318VAC, or 450V. Actually this does not appear as "safe", being on the capacitor voltage limit. But electrolytics can handle an initial surge of up to 10%, so for the first several seconds it takes for tubes to heat up and lower the voltage by drawing current, this is OK for reliable capacitor brands. On the other hand, you have a GZ34 rectifier which will also heat up slowly and not cause much of an initial peak. Still, power tubes may fail etc. etc., so back to a safe value.
The exact voltage calculation is involved and only good for exams for 3rd year students! It is easier to just measure the voltages in practice - as said, you will only be really safe using the peak AC value. Cost-wise there is not much of a saving wrt close spaced voltage capacitors. There are 500V capacitors available which I would prefer here, but perhaps not generally so.
For completeness sake there is also the matter of ripple current that the input capacitor must be able to handle, but that is another topic. If not known to you, it does mean that you could not use just any (physically small) capacitor here even if the capacity is right. (They will heat up and eventually self-destruct.) Specifications are available for most makes.
The exact voltage calculation is involved and only good for exams for 3rd year students! It is easier to just measure the voltages in practice - as said, you will only be really safe using the peak AC value. Cost-wise there is not much of a saving wrt close spaced voltage capacitors. There are 500V capacitors available which I would prefer here, but perhaps not generally so.
For completeness sake there is also the matter of ripple current that the input capacitor must be able to handle, but that is another topic. If not known to you, it does mean that you could not use just any (physically small) capacitor here even if the capacity is right. (They will heat up and eventually self-destruct.) Specifications are available for most makes.
Well, many thanks again. Actually that wasn't quite my questiion, but I'll put that information to good use too. As it turns out, that cap is actually a 500V rated cap. The value on the schematic was just copied from the original.
My poor phrasing obscured my intended question, which was to find a way to calculate the max. cap input for the rectifier, given the somewhat different arrangement of capacitances and resistances in this case. As a practical matter, I'm not too worried. And the various imperfect models of the circuit in PSUD all showed lower transient voltage and current values than the maximum stated in the data sheets. So I can see how the rectifier holds up and modify the circuit if it becomes a problem. Just for advancing my skills, I was looking for a reference on how to calculate the max cap size when adding series resistors and including winding resistance or other elements like the bias cap. The general rectifier data sheets only specify a simple cap input.
Sheldon
My poor phrasing obscured my intended question, which was to find a way to calculate the max. cap input for the rectifier, given the somewhat different arrangement of capacitances and resistances in this case. As a practical matter, I'm not too worried. And the various imperfect models of the circuit in PSUD all showed lower transient voltage and current values than the maximum stated in the data sheets. So I can see how the rectifier holds up and modify the circuit if it becomes a problem. Just for advancing my skills, I was looking for a reference on how to calculate the max cap size when adding series resistors and including winding resistance or other elements like the bias cap. The general rectifier data sheets only specify a simple cap input.
Sheldon
I am beginning to feel sensitive about dominating this discussion; there are others at least as capable as myself to advise here, but just quickly my approach.
As said before, the rectifier is protected by the total series impedance, which limits the peak current. The exact calculation requires inter alia knowledge of the impedance of the transformer winding, which is mostly unavailable. Data sheets give specs. in terms of minimum dc resistance (in serie with each rectifier anode), so one usually relies on that. Often also a maximum value of capacitor is given. But it will be clear to you that one may exceed that C-value so long as there is enough resistance in serie. If you use a higher value of the specified max. C for whatever reason (lower ripple), you will need to use a higher resistance than specified. (It will of course affect the available dc voltage: it will be clear that in the limit you would have gone right across the field to a choke input filter.)
Different data sheets can give somewhat different values. A quick summary: RCA Tube Manual specifies for 450Vrms input an effective serie impedance of 160 ohm but do not specify for what input capacitance. Philips is rather more complete. They specify for 300Vrms 2 x 75 ohm, and for 350Vrms 2 x 100 ohm, and so on to 550Vrms: 2 x 200 ohm. The specified input capacitor is 60uF. They also specify that the resistance is the total source resistance, being Rsecondary + [👎sq. x Rprimary]. You can of course use just a single resistor in the cathode lead.
You will notice that while they are talking (correctly) about an impedance, the specs. refer only to the dc resistance. But the difference might be minimal; I have never measured power transformer impedance. For my money I have relied on these specs. and got on with the job.
Others say......?
As said before, the rectifier is protected by the total series impedance, which limits the peak current. The exact calculation requires inter alia knowledge of the impedance of the transformer winding, which is mostly unavailable. Data sheets give specs. in terms of minimum dc resistance (in serie with each rectifier anode), so one usually relies on that. Often also a maximum value of capacitor is given. But it will be clear to you that one may exceed that C-value so long as there is enough resistance in serie. If you use a higher value of the specified max. C for whatever reason (lower ripple), you will need to use a higher resistance than specified. (It will of course affect the available dc voltage: it will be clear that in the limit you would have gone right across the field to a choke input filter.)
Different data sheets can give somewhat different values. A quick summary: RCA Tube Manual specifies for 450Vrms input an effective serie impedance of 160 ohm but do not specify for what input capacitance. Philips is rather more complete. They specify for 300Vrms 2 x 75 ohm, and for 350Vrms 2 x 100 ohm, and so on to 550Vrms: 2 x 200 ohm. The specified input capacitor is 60uF. They also specify that the resistance is the total source resistance, being Rsecondary + [👎sq. x Rprimary]. You can of course use just a single resistor in the cathode lead.
You will notice that while they are talking (correctly) about an impedance, the specs. refer only to the dc resistance. But the difference might be minimal; I have never measured power transformer impedance. For my money I have relied on these specs. and got on with the job.
Others say......?
- Status
- Not open for further replies.