Huge inrush currents in some Amateur HF Transceivers.
Recently I designed an overvoltage/overcurrent protection device to replace the old-fashioned crowbar circuit. Some interesting conclusions arose during testing of the prototypes. Many Yaesu, Kenwood and Icom transceivers (and probably others) take a significant inrush current on switch-on. These currents can be large enough to activate the shutdown circuitry of the high speed LTC4368 Overvoltage / Overcurrent protection IC. The Kenwoods and Icoms tested so far have been ok but the highest inrush current seen to date is 72 amps with the Yaesu FT991. The inrush current is so great that there is a simultaneous voltage drop of the 13.8V power rail by approximately 50%. Ultra-fast overcurrent protection circuits see this as a potential fault and will close down.
It should be stressed that alarming as it sounds, in itself a high inrus current is not necessarily a problem and this article does not imply criticism. These inrush currents take place in such a short time that it is actually quite difficult to measure. In most cases they are unnoticed and of no concern.
As an intellectual exercise it is worth reviewing the facts around these inrush currents however because of the insight to be gained in modern rig design and component characteristics. It appears that this feature of the Yaesu FT991 has not been discussed (nor even noticed!) elsewhere.
I’ll explore this in two parts, the first one being methodology and results, the second one being a discussion of the design issues facing the mainstream transceiver manufacturers.
Channel One of the ocilloscope measures the voltage developed across a 4 milli-ohm current sense resistor. Running in single shot mode the scope captures the trigger moment when the current starts to rise, coincident with switching on the Yaesu. I=V/R so for example 200mV measured means the current is 0.2 / 0.004 = 50 amps. In tandem with that, Channel Two is used to capture the change in voltage on the main supply rail from the PSU.
At the moment of switch-on the oscilloscope shows an inrush current of 289mV / 4 milliohms = 72 amps for around 30 microseconds. Current remains above 50 amps for 100 full microseconds. The supply rail voltage drops by 5.2 Volts.
Icom IC7300, Kenwood TS590 and TS890S radios all have inrush currents under 30 amps.
We are used to thinking about inrush currents due to the presence of large capacitors in power supply units. You will find lots written on the theme in connection with high power linear amplifiers for example. But analysis of the schematics of several modern 13.8V transceivers reveals no huge capacitors, so what is going on ?
Firstly, the transceivers expect to be fed with a reasonbably smooth 13.8V supply so they do not require large smoothing capacitors. The FT991 however has more than some other rigs. Examining the schematic shows a 1,000uF, a 470uF and several smaller electrolytics. It is always a good idea to use plenty of decoupling on power rails with any electronic device so as expected there are also generous quantities of 100nF ceramics.
An analysis of the schematic is shown at the end**, but the details of all the capacitors on the switched 13V rail were put into LTSpice electronic circuit simulation software. Typical ESR values for the ceramics and good quality low ESR electrolytics were configured. There is a common mode choke for HF RF on the input so this was included (has the effect of slowing down the inrush). It is assumed the choke does not have too high an inductance or there would be a danger of back emf from power failure causing damage. To provide a good comparison with the oscilloscope trace a 4 milli-ohm resistor was put in series with the supply.
The result of the simulation in terms of supply current is show below. It is a remarkably good match to the oscilloscope trace for the first 0.2 milliseconds subsequent to the start of the surge. The oscilloscope curve then flattens out for another 0.2 milliseconds which is a little more than the simulation shows. It is guessed that this is due to the various voltage regulators switching on and supplying current to their own load capacitors – something not modelled yet in the LTSpice simulation.
If we remove the larger 1,000uF capacitor we get a better approximation to the schematic of the Kenwodd TS590 and the Icom 7300 both of which exhibit less than half the inrush peak of the FT991.
So is this poor design ? Is it a problem ? That is more philosopical than electronic – it all depends on your requirement and what the designer is trying to acheive. The important thing is to be aware of these transients and of what their effect may be on your power arrangements, expecially if you are thinking of employing ultra high speed protective devices.
|Update 13/08/2021. After discussing this problem with another Yaesu user it occurred to me that there might be a better solution than just deciding not to use the built-in overcurrent protection. |
BOB will sense and close down in 8 microseconds when built according to the LTC datasheet. But by making the current sense part of the circuit into a kind of low pass filter it is possible to require a full 1 millisecond of overload before protection kicks in. This copes with the 70+ amps power on surge of the FT991 yet allows a 20 amp overload to be set. Any overcurrent more than 20 amps for more than 1 millseconds causes close down. Not as fast as 8 microseconds but a factor of 100-10,000 faster than any fuse. That has to be worth going for.
I’ve simulated it in Spice and it works a treat, exactly as designed. Not yet done a real life test – real life kind of gets in the way 🙂
I’ll document it fully here when I get the time and will also report any real live tests. The modification, by the way, is not easy to do on the existing circuit board. If I ever make another run then I can incorporate it properly.
**A More detailed analysis of FT991 Schematic. follows
This is where the external 13.8V supply arrives via a common mode choke. It has 0.001, 0.01 and 0.1uF capacitors across it for decoupling.
The 13.8V unswitched supply goes off to the control unit from which the Power-On (PON) signal arrives when the power on button is held down by the operator. This signal switches on a relay which then latches. This provides a 13VS (13V-switched) supply which goes to many places. There is a 10uF to ground on the relay contacts.
The 13VS goes to Q4009, Q4010 and Q4011 9V, 5V and 3.3V regulators each of which has a 10uF capacitor on its input together with several smaller ceramics.
The fan control circuitry also becomes active with an unspecified electrolytic on its 13VS rail.
13VS goes directly to a 1,000uF capacitor and through Diode D1068 a 470uF capacitor. It also provides 13VS to a variety of relay coils.
This is permanently powered by the 13VUS rail which feeds three regulators. These do not therefore come into play for inrush when switching on.
Note that Yaesu have named the 13VS rail on the wiring interconnections on this board as 13V. Not nice to have inconsistencies of annotation ! On this unit the 13V(S) rail is distributed to several other pins for interconnection to other boards but it isn’t actually used much and has no electrolytics, just several small ceramic capacitors.
The 13VS has the usual ceramics plus a 47uF capacitor at its input to the board.
13VS feeds a 5V regulator and the control high side switches for all the tuner relays. A handful of ceramics.
PLL, EDSP, Panel, Fan Control, Jack, DSP-1 UNITS
No 13VS to these units.
VDSL and EMC
For some time I have been in the habit of operating at the same time every day on 40m with 100W CW. Chatting to a neighbour I found that they were thinking of moving ISP because the ISP could not fix an intermittent VDSL problem that occurred around the time of my transmissions.
A simple test revealed that true enough, 40M RF caused loss of VDSL connection. An examination of the telephone wiring revelaed the reason why. It was old, very old – there was no NTE-5 of any age just an ancient punch down line jack. A rats nest of extension wiring was connected directly to the line. The VDSL filter was actually an ADSL filter and it was in line with the only telephone in use in the house, a DECT phone.
The current ISP had sent out the modem for the end-user to connect when in fact they should have sent an Openreach engineer to do an install. No Openreach engineer would have permitted installation of VDSL without installing the correct NTE-5 and filters.
Our neighbours are great folk and rather than just doing the minimum I offered to re-install all the phone wiring and bring it up to scratch. They would benefit from an increse in speed, a higher level of reliability and importantly to them I would sort out some very ugly wiring with loops of cable and badly installed wall warts. From my perspective I would have peace of mind knowing that I wasn’t likely to be a source of an EMC problem, even though obviously, since a VDSL shouldn’t respond to environmental RF, I couldn’t actually be held responsible.
I don’t advise others to take this action unless they are qualified to do so. I am pleased to be a professional chartered engineer with many years telecomms industry experience to prop me up and I felt qualified to undertake this role. I also had assessed the level of risk to house infrastructure and I wouldn’t be driling holes, just using surface conduit, removing defunct wiring and attaching a brand new NTE-5 to an existing back box.
To be sure of getting the best results for my neighbour I made a small balanced line rf sniffer and examined the telephone line signal with my Spectrum Analyser, a Siglent SVA1000X series. I could see the VDSL signal clearly at 7Mhz. It was quite a poor level, presumably becasue of the faulty installation and when I put my 40m TX on an autosender I could see the VDSL signal was poor in comparison with 7Mhz RF. A notch filter was designed and made up using a T32-6 toroid and this was tested on the SA and tuned to 7.1Mhz.
The day for the work came around and all surplus wiring was disconnected and tied back, out of sight. The small 40m notch filter was placed across the line on the consumer side of the NTE5. Toroidal cores were used for common mode chokes on the router power line, the VDSL line just prior to entry in the router and on some direct connect Ethernet cables attached to the router. All telephone cabling was reterminated and the system brought back up again.
Tests revealed that the synced VDSL speed had increased significantly and that no amount of RF that I could generate had any impact. An excellent result and no drama.
01 July 2021