On the servers I manage I use the postfix MTA (mail transfer agent). I occasionally find things about it that are not documented. I am adding this post to track those things so I can find them when need them. I hope some others can benefit from my notes.
DKIM
After setting up CWP (Control Web Panel) I found that users could not send email. It was being rejected due to a duplicate DKIM header. Literally the same, exact DKIM header appearing twice in the headers. CWP uses postfix with opendkim. Postfix was calling opendkim twice, both before and after the Amavis anti-virus scan. The fix was a modification to master.cf to skip the milters call (where opendkim lives) after the Amavis call.
127.0.0.1:10025 inet n - y - - smtpd
-o smtpd_milters=
Add the -o smtpd_milters= entry in the 127.0.0.1:10025 inet n - y - - smtpd block of the master.cf configuration file and the duplicate DKIM headers go away.
I have been working on a long term project to convert my (and my Father’s) old reel-to-reel tapes to digital. As part of that I have been working on a Teac X3mkii deck. Skipping all the mechanical fun, at this point it is pulling tape very nicely (for a consumer deck) and I was impressed with the 8 KHz stability coming off of the MRL alignment tape. The purpose of this post to search for an answer to an electrical problem with the deck. The problem is ion both channels, but much more pronounced in the right. There is an about 18KHz tone present. See the following spectrum.
This the right channel, the left also has the problem, but much less so. Here the peak is about 40dB above the noise floor, in the left channel it is only about 15dB above the noise floor. Ignore the spikes of hum at the low frequencies, those are easy to fix in the computer, the spike at about 18KHz is not. It is not a pure tone and it is not constant. See the spectrograph below.
Note that the 18KHz pulses on and off.
I have traced it down to the playback head amplifier, see the schematic below. Thanks to HiFi Engine for the X300 Service Manual (close enough to the X3Mkii).
U101 is a 4558 dual op-amp and I tried replacing it with a 5532 with no change to the problem tone. I also looked at the power supply and replaced C306 the decoupling cap and added a 0.47uF right at pin 8 of the DIP to ground, the power supply noise did go down, but the tone is still there. I tried replacing C105 and C107 thinking that they may have gotten old and tired, no change in the 18KHz tone. I even replaced C102 thinking that perhaps the fake DC reference for pin 5 was getting contaminated, but no change in the 18KHz. I am out of ideas. The tone is not visible at pin 3 of U101 but it is there at pin 7. The FETs (Q101 – Q104) are for changing playback EQ based on the settings of two switches (LH1/LH2 and LH/EE), changing these switches does not change the 18KHz tone, but I am starting to think the FET switches are somehow involved.
Why am I obsessing over this 18KHz, not terribly audible tone? Well, even though this is a consumer deck (1/4 track running at 7.5ips), there is material at this frequency coming off of the tapes. See the following spectrograph of one of the tapes. There is clearly energy above 20KHz and I don’t want to lose it by trying to filter out the 18KHz tone (I have tried). The tone at 36KHz (twice the 18KHz) doesn’t worry me as there is no program material there and I can effectively filter it out.
If you have any ideas what I am running into here, please post a followup comment below.
Full size copies of the images above are located here (much easier to see than the small WordPress images above).
P.S. Thanks to iZotope’s RX for showing me the problem and helping me fix many others.
A while back I was working entertainment tech for a small convention, I was the Sound Designer and had inherited the Power Design (usually done by the Lighting department, but we did not find a Lighting Designer until very late). The event was held at a hotel with a 4-section ballroom, call the sections 1, 2, 3, and 4. The stage was set in section 4 and we had a social event in sections 1 and 2 on Friday night with the airwall between 2 and 3 in place. After the social event the airwall between 2 and 3 was struck and the airwall between 1 and 2 was installed. For the remainder of the weekend events were on the stage in section 4 with audience in sections 2 and 3 and section 1 was the performer Green Room.
Power
We were told that we had one 100A 3-phase feed to use (as per the contract with the hotel). We were also told that there were feeders in the airwall pockets between sections 1 and 2 and sections 3 and 4. Call the pocket between 1 and 2 feeder A and the feeder between 3 and 4 we’ll call C and we’ll leave the label feeder B for the airwall pocket between sections 2 and 3 (which we were told did not have a feeder). We decided to use feeder C (airwall pocket between section 3 and 4, nearest to the stage).
100A of 3-phase 120v/208v power gives us 300A of single phase 120v power, but for safety I derated to 80% or 240A. I built the power budget and came out to just about 230A. I assembled my equipment (PDUs) and cable (cam-lock feeders, as well as NEMA 5-15 extension cords) to handle all our needs assuming we were getting all our power from feeder C near the stage.
Contact with the Enemy
When we arrived on Friday morning we discovered that feeder A and C were not 100A, but only 80A. Feeder B was 100A, but if we used it, then we would end up with a PDU in a bad location due to cable lengths. Feeder A and C both came from a power panel in the kitchen behind the ballroom. Feeder B came from a different panel further back in the kitchen, and that panel also fed the panel that fed A and C. Like this:
Panel 1 -> Feeder B (100A)
Panel 1 -> Panel 2 -> Feeder A and C (80A)
We contacted Hotel Engineering and they apologized for getting us bad information. They offered to let us use as many of the different power feeds that we needed to, at no extra cost, to get the job done.
We were not comfortable using just the 80A feeder C as we were very close to that total draw and rather than use feeder B and have a PDU badly located, and against my better judgement, we decided to split the power between feeder C (for the stage) and feeder B (sound and lighting control points and video production control all at the back of the room). We knew where each feeder came from, we knew how the panels tied together, what could go wrong?
Friday night, after the social event, we started setting up the sound and lighting control riser; the video crew started building video production control on the other side of the room. I was the first to tie the two power feeds together when I plugged the mixer outputs into the snake feeding the FOH amp rack. As soon as the connectors mated I heard a hum out of the mains. I immediately knew what was going on. I called the crew together and told them we had a serious potential difference between safety grounds between feeder B and C. One of the crew grabbed a 100′ 12/3 extension cord and ran it diagonally across the room from feeder C to the sound board. We measured over 100mV between the safety ground pins. We assumed there was a wiring fault somewhere between the panels and feeders.
So what do we do now? The first thought was to use a bonding conductor to tie the two safety grounds together. We tried connecting the ground pin of that extension cord to the ground pin of the power from feeder B. That reduced the hum by about 20 dB, but it was still clearly audible. It was clear that we could not get a low enough impedance connection with the 12 AWG power cords we had with us.
We ended up, at midnight, redesigning the power distribution based on the parts we had on hand. We used feeder B since it could carry our full load. One of the PDUs that we added to the order at the last minute combined with multiple sets of camlock T’s of various genders, enabled us to get power where we needed it, but it was not pretty.
