Perhaps check with primary open, and also with each primary half-winding shorting itself.
Mark, in essence, I would have thought the primary would best be loaded by a standard LISN impedance, but you have gone for a shorted reference for expediency or there is practically no change in resulting snubber values?
Mark, in essence, I would have thought the primary would best be loaded by a standard LISN impedance, but you have gone for a shorted reference for expediency or there is practically no change in resulting snubber values?
Then I guess your ~7 VA transformer (18 x 0.4) resembles the 5VA transformer in QM Figure 17, which I had forgotten about till you reminded me. I guess both of these transformers are wound with enormous numbers of turns of very fine wire, producing an enormous leakage inductance. The 250-600 VA transformers I usually measure, have Lleak 100 times less: a few hundred microHenrys.
So (Cx=10nF, Cs=150nF, Rs=680) ought to do the job very nicely for you.
Maybe one of these days I might revise the QM design note, to suggest a double-check: Measure f_osc with Rs=infinity (removed from socket). Then Rs(gives zeta=1) ought to be about [1/(4*pi*f_osc*Cx)]. In your case, it is.
Quasimodo attempts to measure the secondary leakage inductance. Hagerman (link1) suggests measuring L_sec_leak with primary shorted, and so does Basso (link2). I do it that way too.
If you were really curious, you could put together a SPICE simulation of a transformer with leakage inductance(s), connected to a snubber and to a source of bell-ringing pulses (perhaps an abrupt-recovery rectifier driving a big filter capacitor and a load resistor). Then you could measure the zeta of the snubbed secondary, while varying the source impedance that drives the primary. Does zeta change? If so, does it change a lot? Etc.
Exactly. The value of Rs=650R I arrived at indirectly agrees well enough with your calculation.
I just measured another small transformer, also potted, 2x18V/40mA, single primary. With Cx=10nF & Rs=inf. has a natural freq. of 17.5 kHz, i.e. the same order of magnitude as the one I rported previously.
Tomorrow I'll do a power transformer (approx. 250VA). It's somewhere in the cellar, so I must first dig it out.
BTW I also studied a single RC snubber for the first transformer. I excited it with Cx=4.7pF, got a good response and interesting results. I'll double check the results and post them asap.
Thank you very much for your interest in my results.
I just measured another small transformer, also potted, 2x18V/40mA, single primary. With Cx=10nF & Rs=inf. has a natural freq. of 17.5 kHz, i.e. the same order of magnitude as the one I rported previously.
Tomorrow I'll do a power transformer (approx. 250VA). It's somewhere in the cellar, so I must first dig it out.
BTW I also studied a single RC snubber for the first transformer. I excited it with Cx=4.7pF, got a good response and interesting results. I'll double check the results and post them asap.
Thank you very much for your interest in my results.
The mains typically presents a non-zero impedance to the primary winding. The impedance is obviously low at the power line frequency, but appears to generically increase with nominal level of about 1 ohm at 10kHz, rising to about 20 ohm at 100kHz, and levelling around 50 ohm in to the MHz (from curves I've seen - just looking at one in a book by Mark Nave) - which is the basis for standard LISN impedance.
Hagerman's use of a short was obviously to simplify analysis - I don't have Basso's book - and I can appreciate that standardising on a short is the most appropriate way to reference and compare measurements between parts.
Hagerman's use of a short was obviously to simplify analysis - I don't have Basso's book - and I can appreciate that standardising on a short is the most appropriate way to reference and compare measurements between parts.
Not quite. The problem with dialling Rs lower and lower and lower, is that you eventually begin to violate your promise that Rs will be much larger than (the magnitude of impedance of Cs) at the resonant frequency. See the bottom half of p.17. Remember that Cs's only purpose is to minimize power dissipation in Rs. Cs accomplishes this by having an impedance magnitude much larger than Rs at the mains frequency, and by having an impedance magnitude much smaller than Rs at the resonant frequency.
If you are eager to experiment with very large damping factors (thus very small values of Rs) on Quasimodo, I suggest that you decrease the impedance of Cs quite dramatically. This means: increase the capacitance of Cs quite dramatically. Just to be extra-certain that |1/(j*2*pi*f*Cs)| is very low, I recommend increasing its value to (1500 * Cx). If you're using Cx=0.01uF, and if you want to horse around with very low values of zeta for grins and giggles, use Cs = 15uF. You'll need a non-polar capacitor with suitable voltage rating, perhaps like (three of these in parallel).
Edit- by the way, you could do this in SPICE before doing it on your physical Quasimodo. Post#1 of the CheapoModo thread includes an LTSPICE .asc circuit file that might be a nice starting point.
Last edited:
Mark, I really do not think that we have a fundamental disagreement here; it is merely that we interpret the system (transformer secondary with its intrinsic and added parameters) from two points of view. I'm looking at it from the standppoint of the linear system analysis in order to obtain insight into what motions are possible, disregarding thus at first the purpose of the circuit. So we have here an oscillatory system with several constant parameters (LT, CT, RT, Cx, Cs) and one that can be varied (RV1). The latter can assume any value between zero and infinity; and the first question to be answered is what happens in the system at these limit values of RV1. We can easily agree that at RV1=0 the system will oscillate at the natural frequency determined by Cx and Cs in parallel; and at RV1=inf. the natural frequency will be determined by Cx alone. Both oscillations must be damped in character due to the presence of RT. We are tacitly assuming that there is excitation for both modes. Now the only element that can activate the one or the other mode is RV1; and if it were a switch, then only these two states were realizable. Since it is a variable resistance, then an infinite number of states between the two limit ones are possible, the character of which were to be determined by a more detailed modelling and/or measurements; and thanks to this fine jig of yours we are able to study the circuit behaviour at our leisure.
Starting now with the largest resistance value of RV1 chosen (1K in your case, 5K in mine) and reducing it slowly we see that the amplitude of the (unwanted) damped oscillatory motion with the frequency determined by Cx decreases, reaching the state of optimal damping (Zeta=0.707), at which point we stop, having reached our objective.
But if we, for the sake of the experiment, proceed reducing the value of RV1, the first natural mode will be still further damped. However, eventually a damped oscillation of the second natural mode (defined by Cx & Cs in parallel) emerges because at lower RV1 values Cs can partake in the energy exchange between LT and the capacitances present, becomes stronger until, at RV1=0, its amplitude is only governed by RT. This is the situation you clearly warned against in an earlier post.
Neither of these two oscillations are desired, but both are possible and, having comparable amplitudes, both are equally dangerous. Your choice of approaching the point of optimal damping from above (in terms of the RV1 value) is a sensible one in that the purpose of adding Cx is to detune the LT/CT system such that it becomes insensitive to the rectifier capacitances; thus it is this mode that is to be damped. Nevertheles, the second mode is still there, and I would recommend checking the distance (in terms of the RV1 value) between the two in order to assess possible adverse effects of component tolerances. In my case this distance was large enough.
It is on the basis of the above analysis that I reached the conclusion that RV1 positions the operating point of the transformer secondary between two possible natural oscillatory modes.
One last comment: I have no interest in extra low RV1 values; this seems to be a misundestanding.
Please excuse the long post, but I felt the need to explain my logic in some detail.
Starting now with the largest resistance value of RV1 chosen (1K in your case, 5K in mine) and reducing it slowly we see that the amplitude of the (unwanted) damped oscillatory motion with the frequency determined by Cx decreases, reaching the state of optimal damping (Zeta=0.707), at which point we stop, having reached our objective.
But if we, for the sake of the experiment, proceed reducing the value of RV1, the first natural mode will be still further damped. However, eventually a damped oscillation of the second natural mode (defined by Cx & Cs in parallel) emerges because at lower RV1 values Cs can partake in the energy exchange between LT and the capacitances present, becomes stronger until, at RV1=0, its amplitude is only governed by RT. This is the situation you clearly warned against in an earlier post.
Neither of these two oscillations are desired, but both are possible and, having comparable amplitudes, both are equally dangerous. Your choice of approaching the point of optimal damping from above (in terms of the RV1 value) is a sensible one in that the purpose of adding Cx is to detune the LT/CT system such that it becomes insensitive to the rectifier capacitances; thus it is this mode that is to be damped. Nevertheles, the second mode is still there, and I would recommend checking the distance (in terms of the RV1 value) between the two in order to assess possible adverse effects of component tolerances. In my case this distance was large enough.
It is on the basis of the above analysis that I reached the conclusion that RV1 positions the operating point of the transformer secondary between two possible natural oscillatory modes.
One last comment: I have no interest in extra low RV1 values; this seems to be a misundestanding.
Please excuse the long post, but I felt the need to explain my logic in some detail.
If you're nervous that you might not succeed in finding an Rs value which places you in the comfortably safe region (0.5 < zeta < 2.0), you can always increase Cs. To obtain a factor of 3 more breathing-room, increase Cs by a factor of about (3^2)=9.
Try it in simulation*; you'll see that bigger Cs lets you dial down Rs a lot farther before oscillatory ringing reappears. Of course it's still a very bad idea to set Rs=0, even with huge Cs; you do need something to dissipate the energy in the resonant circuit.
You may even be able to find a (Cs/Cx) ratio that provides adequate safety margin on both sides, when the user mistakenly dials Rs to the FlatLine condition** shown by member mcandmar in post#28 of the CheapoModo thread.
*Or try it on a real Quasimodo
**I don't recommend choosing the FlatLine condition; its zeta is not 1.00
Try it in simulation*; you'll see that bigger Cs lets you dial down Rs a lot farther before oscillatory ringing reappears. Of course it's still a very bad idea to set Rs=0, even with huge Cs; you do need something to dissipate the energy in the resonant circuit.
You may even be able to find a (Cs/Cx) ratio that provides adequate safety margin on both sides, when the user mistakenly dials Rs to the FlatLine condition** shown by member mcandmar in post#28 of the CheapoModo thread.
*Or try it on a real Quasimodo
**I don't recommend choosing the FlatLine condition; its zeta is not 1.00
Last edited:
Thank you for the tips.
In the meantime I played some more with Quasimodo. This particular transformer goes into the power supply snubbed with the values already mentioned in my first post and your reply to it, i.e. 10nF/150nF/670R. This combination produces a mild overshoot and then goes down to zero. I'm quite comfortable with this combination.
In the meantime I played some more with Quasimodo. This particular transformer goes into the power supply snubbed with the values already mentioned in my first post and your reply to it, i.e. 10nF/150nF/670R. This combination produces a mild overshoot and then goes down to zero. I'm quite comfortable with this combination.
Quasimodo
It's me who sent a PM to Mark.
I include 2 pics about my situation:
Pic "500VA-1200VA" should give an impression between a standard 500VA/2x42VAC and my special 1200VA/2x28VAC-tranny. Of course my tranny is oversized, but a very low noise design specified for only 1000VA; I assume its core to be at least 1200VA.
Pic "Quasimodo Results" show traces on an old (like me) analgue scope before and after suggested 10nF-series cap. Upper trace (before = input) has 10V/div., lower trace (after = output) has only 5mV/div (!!!).
Results with 500VA-tranny were just normal, identical to Mark's description in his technical article.
It's me who sent a PM to Mark.
I include 2 pics about my situation:
Pic "500VA-1200VA" should give an impression between a standard 500VA/2x42VAC and my special 1200VA/2x28VAC-tranny. Of course my tranny is oversized, but a very low noise design specified for only 1000VA; I assume its core to be at least 1200VA.
Pic "Quasimodo Results" show traces on an old (like me) analgue scope before and after suggested 10nF-series cap. Upper trace (before = input) has 10V/div., lower trace (after = output) has only 5mV/div (!!!).
Results with 500VA-tranny were just normal, identical to Mark's description in his technical article.
Hm4nine
I tested my1KVA Plitron transformers, 18 of them with 2 x 50 volt secondary and dual primary wired for 110. You can read an exchange with Mark from post 261 on. I'm probably not of any help as I was not able to get pictures and my case is outside of the norm due to the size of capacitors . I did obtain a trace that looks very much like Marks critically damped trace.
I tested my1KVA Plitron transformers, 18 of them with 2 x 50 volt secondary and dual primary wired for 110. You can read an exchange with Mark from post 261 on. I'm probably not of any help as I was not able to get pictures and my case is outside of the norm due to the size of capacitors . I did obtain a trace that looks very much like Marks critically damped trace.
hm4nine, your analog scope photo might have the bottom trace inverted (?) Quasimodo uses a state-of-the-art, super low-Rds MOSFET to yank down the left end of the series capacitor Cx, and yank it down HARD. This also yanks down the right end of Cx.
So we expect to see falling-edge waveforms in both traces. But your traces show a falling input and a rising output? That's perplexing.
So we expect to see falling-edge waveforms in both traces. But your traces show a falling input and a rising output? That's perplexing.
Attachments
Quasimodo
Hi Mark!
Thank you for your hint; unfortunately I cannot verify any fault, because in the meantime I used my scope for other measurements.
Nevertheless I rechecked my measurements with Quasimodo; I include some pics about them:
Quasimodo1:
No loads, shows signal before and after 10múF-cap.
References are 5V/div. (both channels) and 1ms/div. Supply is 12VDC.
Test frequency was 120Hz.
Quasimodo2:
With 1200VA-tranny (winding has 28VAC), but other windings left opened and NO damping resistor Rs.
References are 0,1V/div. for 2. channel and 0,1ms/div.
Quasimodo3:
Identical to 2), but now Rs = 1kOhm
Quasimodo4:
Now all unused windings are shortcut, again NO Rs
References are 5mV/div (!!) for 2. channel and 2ms/div
Quasimodo5:
Identical to 4), but now Rs = 316R.
Confusing for me is curve shape in 4) and 5).
Unfortunately your pics in your description don*t show one complete cycle.
Hi Mark!
Thank you for your hint; unfortunately I cannot verify any fault, because in the meantime I used my scope for other measurements.
Nevertheless I rechecked my measurements with Quasimodo; I include some pics about them:
Quasimodo1:
No loads, shows signal before and after 10múF-cap.
References are 5V/div. (both channels) and 1ms/div. Supply is 12VDC.
Test frequency was 120Hz.
Quasimodo2:
With 1200VA-tranny (winding has 28VAC), but other windings left opened and NO damping resistor Rs.
References are 0,1V/div. for 2. channel and 0,1ms/div.
Quasimodo3:
Identical to 2), but now Rs = 1kOhm
Quasimodo4:
Now all unused windings are shortcut, again NO Rs
References are 5mV/div (!!) for 2. channel and 2ms/div
Quasimodo5:
Identical to 4), but now Rs = 316R.
Confusing for me is curve shape in 4) and 5).
Unfortunately your pics in your description don*t show one complete cycle.
pics 2 & 3 show the ringing after the overshoot.
I think these are with the other windings open circuit.
Pic 5 shows the tiny voltages when all the other windings are shorted.
You have to zoom in on the little blips.
There you will see the overshoot and ringing but at a much lower level.
That is the one you can monitor as you change the VR. It rings badly when far too big and rings badly when set to zero R.
In between you can identify the minimum ringing.
I think these are with the other windings open circuit.
Pic 5 shows the tiny voltages when all the other windings are shorted.
You have to zoom in on the little blips.
There you will see the overshoot and ringing but at a much lower level.
That is the one you can monitor as you change the VR. It rings badly when far too big and rings badly when set to zero R.
In between you can identify the minimum ringing.
hm4nine, the Big Quasimodo Idea is shown in the design note Fig.2 (page 3), which I have copied below. Switch SW1 is the bell ringer; the bell is struck when SW1 closes. That's where all the action is: at the instant when SW1 closes, a very steep falling edge is driven into the transformer's secondary leakage inductance LT, and this rings the bell. So we use a dual trace oscilloscope to trigger on the falling edge and to zoom in to the ring-a-ding-ding oscillations that occur immediately after.
Transformer secondary inductance LT is generally between 400 nanoHenrys and 20 milliHenrys. You can grind through the math to see for yourself that, with a Cx of 10 nanofarads, the transformer secondary + Cx will oscillate at a frequency between 12 kilohertz and 2 Megahertz. To see 3-5 full periods of oscillation, the horizontal sweep setting on your scope needs to be between 0.2 microseconds/div and 50 microseconds/div.
Let's have a look at the sweep rates for published scope photos of Quasimodo oscillatory ringing. It's easy to do this because digital oscilloscopes display the horizontal sweep rate right on the scope image itself. Our eyes tell us:
Please have a look at Fig.13 of the QM design note, to see how to correctly test a center tapped transformer (and/or four other kinds of transformer!) with Quasimodo.
Good luck!
Transformer secondary inductance LT is generally between 400 nanoHenrys and 20 milliHenrys. You can grind through the math to see for yourself that, with a Cx of 10 nanofarads, the transformer secondary + Cx will oscillate at a frequency between 12 kilohertz and 2 Megahertz. To see 3-5 full periods of oscillation, the horizontal sweep setting on your scope needs to be between 0.2 microseconds/div and 50 microseconds/div.
Let's have a look at the sweep rates for published scope photos of Quasimodo oscillatory ringing. It's easy to do this because digital oscilloscopes display the horizontal sweep rate right on the scope image itself. Our eyes tell us:
- 2 us/div ...... QM design note Fig 1
- 20 us/div ...... QM design note Fig 3
- 2 us/div ...... QM design note Fig 10
- 2 us/div ...... QM design note Fig 11
- 0.5 us/div ...... QM design note Fig 14
- 10 us/div ...... this diyAudio Quasimodo thread, post #143
- 0.5 us/div ...... this QM thread, post #255
- 20 us/div ...... this QM thread, post #314
Please have a look at Fig.13 of the QM design note, to see how to correctly test a center tapped transformer (and/or four other kinds of transformer!) with Quasimodo.
- Be sure to set your scope vertical amplifiers, to DC coupled {not AC coupled}
- Be sure to switch OFF your scope's Bandwidth Limit feature
Good luck!
Attachments
I think I understood the idea; and of course I read your design notes including how to connect different kinds of trannies. My tranny has 2 seperated secondaries and one primary. Therefore both one secondary and primary were shortcut.
But which I obviously did not consider was small hotizontal sweep rate.
Scope now was DC-coupled (though AC-coupling made no difference), timebase was set to 0,2usec and input sensivity to 2mV. These are the limits of my old analogue scope....
To make a long story short: Rs now must be approx. 12,7R - and I really should look for a new scope.....
Thanks for your help!
But which I obviously did not consider was small hotizontal sweep rate.
Scope now was DC-coupled (though AC-coupling made no difference), timebase was set to 0,2usec and input sensivity to 2mV. These are the limits of my old analogue scope....
To make a long story short: Rs now must be approx. 12,7R - and I really should look for a new scope.....
Thanks for your help!
Wow! Readers may wish to consult post #89 in this thread, written back in November 2013, to get a five-month-old datapoint of at-cost pricing :
But you should keep in mind that I required payment by mailing a physical check or mailing currency (banknotes): no wire transfer, no PayPal, no Bitcoin. Maybe the convenience of these newer payment options, adds another layer of complexity and therefore another layer of costs.
Last edited: