"legg" wrote in message news:hkqgpa113smdh8hgghccohq3m20l2j2lvh@4ax.com...
On Sat, 4 Jul 2015 15:27:30 -0700, "Harry D" <harryd@tdsystems.org>
wrote:
><snip>
>One wonders what fixed-frequency switch currents might be generated at
>high-line, using components sized for low-line power delivery, without
>ever disturbing the low frequency line current being monitored by the
>control chip.
>
>As I pointed out, this is just one step further of decoupling the
>switch current stress from the currents actually being monitored and
>limited by the controller being used here.
>
>RL
>Not sure of the point that you are trying to make.
>Fig. #8 top waveform is the current in the RCI's AC winding and Fig. #9 is
>the current in the DC winding. As shown by the photos the unit is operating
>close to CrCM. Not by design but just turned out that way. If the switch
>turned on 500nS sooner we would be in true CrCM. The original T.I. inductor
>was 1203 uH and the RCI inductance was reduced to 393 uH to increase the
>ripple current and show off it's performance. It just so happened that the
>load selected placed the unit close to CrCM operation.
>CrCM operation causes the highest ripple current and since they are
>variable
>frequency devices, cannot be easily interleaved. As such, this high ripple
>current causes large input filters and huge X caps. The dreaded X cap
>circulate out of phase current and piss off the power companies and also
>corrupts the PF. As power is increased, on time increases and the frequency
>drops. If the RCI is tuned for max rejection at the CrCM lowest frequency
>(>40 dB) then you have a nice match. As the load decreases, the frequency
>increases an the ripple current decreases but you still have >20 dB
>rejection from the RCI. This should enable higher power CrCM units to be
>viable products.
>
>Harry
>
It's an enticing concept; are you sure resonance is a critical
component of the technique? Isn't it used already for static DC-DC
anywhere?
Did you do any fiddling around with the resonant winding ~isolated?
There may be an unfortunate claimant to the acronym, but I wouldn't
let that bother me.
RL
RL
Resonance is not necessary but it makes sense to keep Cr as small as
possible because at the line frequency rectified (120 Hz) it is in parallel
with Cx and we do not want it to draw reactive line current.
It is strange that this concept has not been used in the past, maybe this
will get the ball rolling. There are old papers but they try to explain it's
operation using transformer theory making it totally confusing. Just a
coupled inductor with a weak K factor causing two more leakage inductors to
form in very advantageous places.
Thanks,
Harry
Reply by legg●July 4, 20152015-07-04
On Sat, 4 Jul 2015 15:27:30 -0700, "Harry D" <harryd@tdsystems.org>
wrote:
><snip>
>One wonders what fixed-frequency switch currents might be generated at
>high-line, using components sized for low-line power delivery, without
>ever disturbing the low frequency line current being monitored by the
>control chip.
>
>As I pointed out, this is just one step further of decoupling the
>switch current stress from the currents actually being monitored and
>limited by the controller being used here.
>
>RL
>Not sure of the point that you are trying to make.
>Fig. #8 top waveform is the current in the RCI's AC winding and Fig. #9 is
>the current in the DC winding. As shown by the photos the unit is operating
>close to CrCM. Not by design but just turned out that way. If the switch
>turned on 500nS sooner we would be in true CrCM. The original T.I. inductor
>was 1203 uH and the RCI inductance was reduced to 393 uH to increase the
>ripple current and show off it's performance. It just so happened that the
>load selected placed the unit close to CrCM operation.
>CrCM operation causes the highest ripple current and since they are variable
>frequency devices, cannot be easily interleaved. As such, this high ripple
>current causes large input filters and huge X caps. The dreaded X cap
>circulate out of phase current and piss off the power companies and also
>corrupts the PF. As power is increased, on time increases and the frequency
>drops. If the RCI is tuned for max rejection at the CrCM lowest frequency
>(>40 dB) then you have a nice match. As the load decreases, the frequency
>increases an the ripple current decreases but you still have >20 dB
>rejection from the RCI. This should enable higher power CrCM units to be
>viable products.
>
>Harry
>
It's an enticing concept; are you sure resonance is a critical
component of the technique? Isn't it used already for static DC-DC
anywhere?
Did you do any fiddling around with the resonant winding ~isolated?
There may be an unfortunate claimant to the acronym, but I wouldn't
let that bother me.
RL
RL
Reply by Harry D●July 4, 20152015-07-04
"legg" wrote in message news:8jhgpal8r3tnnpios7pq86dib526ba5m4g@4ax.com...
On Sat, 4 Jul 2015 11:05:43 -0700, "Harry D" <harryd@tdsystems.org>
wrote:
<snip>
>What you have done is to remove actual switch current one step farther
>away from the current being monitored by the controller. This makes
>the peak line current limiting method marginally less effective in
>providing some kind of protection for the switch.
>
>With a ~limit on switch dI/dT, perhaps this isn't really an issue.
>
>RL
>
>You bring up lot's of good points. Let's save the CrCM issue until later
>and
>hit the "current being monitored by the controller" issue. The RCI does not
>change in any way how the current is being monitored because the resonate
>capacitor is tied directly back to the bridge return and is not in the
>current sense loop. The current sense waveform is not changed in any way.
>It
>contains both Ac and Dc components of the RCI. The controller's mode of
>operation is not effected by using a RCI.
>
>Regards,
>Harry
>
>
Is figure 8 showing the drain current of the switch, or the current in
the second winding? I suspect the latter, as the drain voltage
indicates that current is still being delivered to the boost
rectifier.
For some period after the boost rectifier turns off, the inductor
winding currents must sum to an approximate negative value, allowing
the drain voltage to fall in a resonant transition not associated with
the conversion or resonant frequencies (in figure8).
At this specific phase angle of the input voltage, and under the
specific loading, a near-zvt and near zct turn-on is demonstrated.
The 100KHz modulator turns the switch off when the peak switch current
is 2xIin +x.......Will this always be the case at turn-on and turn-off
for the switch?
If that is the case, then limiting the peak low-frequency input
current (as is done in the UCC3817) might be reasonably expected to
perform a fair limiting service to the switch.
On the other hand, if the 100KHz modulator could demand larger duty
cycles in order to correct for a load transient or sudden line voltage
surge, there would be no guarantee that the current would not rise to
a larger value, limited only by the resonant capacitor's ability to
deliver it. As this capacitor must be sized to handle the rated load
at the fixed conversion frequency, you could expect 2xIpk of a full
load low-line condition (+n).
One wonders what fixed-frequency switch currents might be generated at
high-line, using components sized for low-line power delivery, without
ever disturbing the low frequency line current being monitored by the
control chip.
As I pointed out, this is just one step further of decoupling the
switch current stress from the currents actually being monitored and
limited by the controller being used here.
RL
Not sure of the point that you are trying to make.
Fig. #8 top waveform is the current in the RCI's AC winding and Fig. #9 is
the current in the DC winding. As shown by the photos the unit is operating
close to CrCM. Not by design but just turned out that way. If the switch
turned on 500nS sooner we would be in true CrCM. The original T.I. inductor
was 1203 uH and the RCI inductance was reduced to 393 uH to increase the
ripple current and show off it's performance. It just so happened that the
load selected placed the unit close to CrCM operation.
CrCM operation causes the highest ripple current and since they are variable
frequency devices, cannot be easily interleaved. As such, this high ripple
current causes large input filters and huge X caps. The dreaded X cap
circulate out of phase current and piss off the power companies and also
corrupts the PF. As power is increased, on time increases and the frequency
drops. If the RCI is tuned for max rejection at the CrCM lowest frequency
(>40 dB) then you have a nice match. As the load decreases, the frequency
increases an the ripple current decreases but you still have >20 dB
rejection from the RCI. This should enable higher power CrCM units to be
viable products.
Harry
Reply by legg●July 4, 20152015-07-04
On Sat, 4 Jul 2015 11:05:43 -0700, "Harry D" <harryd@tdsystems.org>
wrote:
<snip>
>What you have done is to remove actual switch current one step farther
>away from the current being monitored by the controller. This makes
>the peak line current limiting method marginally less effective in
>providing some kind of protection for the switch.
>
>With a ~limit on switch dI/dT, perhaps this isn't really an issue.
>
>RL
>
>You bring up lot's of good points. Let's save the CrCM issue until later and
>hit the "current being monitored by the controller" issue. The RCI does not
>change in any way how the current is being monitored because the resonate
>capacitor is tied directly back to the bridge return and is not in the
>current sense loop. The current sense waveform is not changed in any way. It
>contains both Ac and Dc components of the RCI. The controller's mode of
>operation is not effected by using a RCI.
>
>Regards,
>Harry
>
>
Is figure 8 showing the drain current of the switch, or the current in
the second winding? I suspect the latter, as the drain voltage
indicates that current is still being delivered to the boost
rectifier.
For some period after the boost rectifier turns off, the inductor
winding currents must sum to an approximate negative value, allowing
the drain voltage to fall in a resonant transition not associated with
the conversion or resonant frequencies (in figure8).
At this specific phase angle of the input voltage, and under the
specific loading, a near-zvt and near zct turn-on is demonstrated.
The 100KHz modulator turns the switch off when the peak switch current
is 2xIin +x.......Will this always be the case at turn-on and turn-off
for the switch?
If that is the case, then limiting the peak low-frequency input
current (as is done in the UCC3817) might be reasonably expected to
perform a fair limiting service to the switch.
On the other hand, if the 100KHz modulator could demand larger duty
cycles in order to correct for a load transient or sudden line voltage
surge, there would be no guarantee that the current would not rise to
a larger value, limited only by the resonant capacitor's ability to
deliver it. As this capacitor must be sized to handle the rated load
at the fixed conversion frequency, you could expect 2xIpk of a full
load low-line condition (+n).
One wonders what fixed-frequency switch currents might be generated at
high-line, using components sized for low-line power delivery, without
ever disturbing the low frequency line current being monitored by the
control chip.
As I pointed out, this is just one step further of decoupling the
switch current stress from the currents actually being monitored and
limited by the controller being used here.
RL
Reply by Harry D●July 4, 20152015-07-04
"legg" wrote in message news:uf1gpadh88l8uj6peom9dqrrdmaqnm4l8r@4ax.com...
On Fri, 3 Jul 2015 17:49:06 -0700, "Harry D" <harryd@tdsystems.org>
wrote:
Harry,
At first glance, use in CrCM might be a stretch, due to the switching
frequency variation required to maintain that condition....
As the UCC3817 is a fixed frequency, average current mode controller,
the occurence of CrCM will be incidental at some low frequency phase
angles and loading, rather than being deliberately enforced, on your
demo board. Continuous or discontinuous HF switching modes are
similarly ignored.
The lack of distortion is solely due to the average current mode
control method.
What you have done is to remove actual switch current one step farther
away from the current being monitored by the controller. This makes
the peak line current limiting method marginally less effective in
providing some kind of protection for the switch.
With a ~limit on switch dI/dT, perhaps this isn't really an issue.
RL
You bring up lot's of good points. Let's save the CrCM issue until later and
hit the "current being monitored by the controller" issue. The RCI does not
change in any way how the current is being monitored because the resonate
capacitor is tied directly back to the bridge return and is not in the
current sense loop. The current sense waveform is not changed in any way. It
contains both Ac and Dc components of the RCI. The controller's mode of
operation is not effected by using a RCI.
Regards,
Harry
Reply by legg●July 4, 20152015-07-04
On Fri, 3 Jul 2015 17:49:06 -0700, "Harry D" <harryd@tdsystems.org>
wrote:
Harry,
At first glance, use in CrCM might be a stretch, due to the switching
frequency variation required to maintain that condition....
As the UCC3817 is a fixed frequency, average current mode controller,
the occurence of CrCM will be incidental at some low frequency phase
angles and loading, rather than being deliberately enforced, on your
demo board. Continuous or discontinuous HF switching modes are
similarly ignored.
The lack of distortion is solely due to the average current mode
control method.
What you have done is to remove actual switch current one step farther
away from the current being monitored by the controller. This makes
the peak line current limiting method marginally less effective in
providing some kind of protection for the switch.
With a ~limit on switch dI/dT, perhaps this isn't really an issue.
RL