The Ampex AG-440C Journal

The following journal entries were originally posted to

Part 1, posted 5/23/01

I've been bring to life an AG-440C, and thought it would be good to share my experiences doing it. I've also got some questions, too. I've already made some good progress, so the first installments of this journal will be based on the work done over the last two weeks.

About two years ago I picked up a pair of AG-440Cs from the son of a friend of mine. He had bought them at auction from North Texas State University (Denton, Texas). One was in a roll-around cart, the other had no cart. These were stereo units, one was 1/4-track, the other 1/2-track. Overall, these two units seemed to have moderate use - not creme-puffs, but not run into the ground. One pair of electronics was beat-up in handling (including a broken VU meter). The plan is to bring one unit up in good condition, and use the other as a parts unit. My goal is to have a good-quality 1/2-track record/playback system for playing back archive tapes, and copying tapes.

May 14-15, 2001 -

Cleaned the dust off of both AG-440s. Except for the damaged electronics, both units seem to be in good shape. Both transports have synchronous motors and both have a "Tentrol" looking tachometer wheel on the supply motor.

After cleaning up the cosmetics of the transport in the cart, I was ready to turn it on. However, I needed to oil the capstan and idler. A quick trip to the Ace Hardware in Yerington did not turn up any turbine oil. However, my friend Jerry, who does appliance repair, swore that he bought turbine oil at the Ace Hardware. I went back, and after looking around, found a nice 4-oz. bottle called "Zoom-Spout". On the back it said "Highest quality turbine all-purpose lubricating oil (certified)", so this looked like the right stuff. Plus, the "zoom-spout" makes it easy to get a few drops of oil into tight places. Back home, the oil was applied to the top capstan bearing and inside the idler wheel, per Ampex instructions.

Powering up the transport revealed no unusual bearing noises. It handled tape just fine, so, since I don't have the spring scales yet, I will do the tension calibration later.

I then turned to the clean electronics units. Putting them in the rack above the transport and hooking them up to the transport and my Audio Precision System 2, I ran some signals through the record input to playback line output. Yuck! There was almost no low frequency response, and there were tremendous variations in the gain between different reproduce and record plug-in modules. Most of the low frequency loss seemed to be in the reproduce line amplifier stage. I had bought some capacitors to do recapping, but at the moment, all capacitors were original. I recapped the tubular Mallory electrolytics in one line amp stage, and things got much better.

I took the old capacitors, as well as some of the tantalums, over to the ESI impedance bridge. At first I thought something was wrong with the bridge, but it tested new caps just fine. The Philips replacement caps (obtained from Digi-Key) were pretty close to their spec'd value and had dissipation factors of about 0.02 to 0.1. I was finally able to null the bridge on the old electrolytic caps, and they typically had a capacitance of 0.02uF and a dissipation factor of 3 and higher! All the Mallory electrolytics I had removed were similarly dysfunctional. Just as a double check, I pulled out a General Radio 1650A bridge, and got similar readings on the capacitors. The little epoxy-drop tantalums were much better - nearly right-on in value and with dissipation factors of from 0.05 to 0.3. New tantalums typically have lower dissipation factors (nearly always less than 0.1), so there was a bit of degradation, but nowhere as bad as the electrolytics. The capacitors in question were all Mallory types with date codes that ranged from 7251A to 7418G. I had never seen 27 year-old electrolytics go so completely bad, especially with no sign of heating or leakage. I did notice that the clear plastic shrink wrap on the capacitors was slightly whitened around the edges. Could an inappropriate flux or board cleaner have been used on these?

Next part: after the recapping

- John Atwood

P.S. - Please make comments on what I have been doing. Am I missing things or doing things wrong? I do not have the manuals or schematic for the AG-400C, although I have the AG-440(A) schematic and alignment info.

Part 2, originally posted 5/24/01, later posted as parts 2.1 and 2.2 on 5/27/01

Thanks to all of you who have replied so far. One of the first things I did before powering up the electronics was to clean the fingers on all the PC boards (including the ones in the transport), and apply DeOxit to them. I also replaced the two electrolytic capacitors I could find in the electronics: the 250uF output capacitor on the left side panel, and the 4uF capacitor at the octal record input selector.

I actually had come across one other case before of hopelessly bad electrolytics - when repairing a Tektronix 465B oscilloscope. The power supply filter caps were PC-mount "twist-lock" types, and two out of the three were very bad. Interestingly, they were Mallory, and I think the date code was around 1974! Clearly Mallory had a problem about this time. Most older solid-state Ampex equipment I've seen used Sprague "Littl-lytics", which seem to hold up pretty well.

One other thing I forgot to mention. The scrape idler was pretty squeaky. I didn't have the equipment to do the correct ultrasonic cleaning and lubricating, but tried putting a drop of "Break-Free" lubricant (often used to lubricate guns) in each bearing, cleaning off the excess. The idler runs very smoothly now.

By the way, here are some pictures: The whole machine: and the bad capacitors:

May 16 - 18, 2001:

Before starting this project, I realized that I had to decide on a standard recording tape and recording level. Based on comments on the Ampex mail list, it looked like BASF would be better than Quantegy. Going to the BASF/Emtec web site and checking out their specs, I decided to go with BASF 911 tape, since most of the recordings would likely end up as archives. Based on Jay McKnight's excellent publication "Choo&U" (available on the MRL website), it looks like 355 nWb/m would be good (I have no plans for noise reduction). Since I'm in America, I will start with NAB equalization (especially since virtually all the archive tapes I have are NAB). I then ordered three test tapes from MRL: 233-423-450-105 (7.5ips NAB 180nWb/m), 243-423-510-101 (15ips, NAB, 355nW/m), and 241-570-480-104 (15ips 3150 Hz flutter and speed test tape). The first two tapes are specially made to be used with the Audio Precision System 1, and have no voices during the frequency sweeps.

I also had an old 7.5ips MRL conventional test tape from 1983, when I was working on an old Ampex 300. I would do preliminary work using this tape (for 7.5ips), then use the Audio Precision tapes when things are pretty much working. However, when I put the old test tape on the machine, guess what? Sticky-shed! It wasn't too bad, and most of what was shedding was the back-coating. Putting it into a dish-warming oven for 24 hours seemed to bring it back to normal. With the dry desert air here, it should hopefully stay good for a while.

I searched for various dealers for BASF tape. The one in Reno was essentially useless, so after poking around on the internet and on the phone, ended up ordering a case of ten 10.5 reels of 911 tape for $21/reel from Comtel in Burbank, Calif. Is this a good price, and are there better places to buy?

Most of the lamps were burned-out, so I ordered new lamps from Mouser Electronics. Their 606-CM120MB fits the 120V power-on lamp on the transport and their 606-CM1828 fits the lamp over the VU meters. I was unable to find replacements for the yellow 28V Dialco plug-in "Ready" lamps, but scavenged a good one from one of the extra electronics.

Back to the electronics:

I had ordered a bunch of replacement electrolytics from Digi-Key. I planned to replace the little tantalums with high-quality Panasonic radial capacitors, type KS and KG ("a substitute for tantalum in many applications"). Sadly, Panasonic no longer makes high-quality axial types, so I ended up getting the Beyschlag-Centralab type ASM-021 (marked "Philips") from Digi-Key. The electrolytics and tantalums in two channels worth of electronics were replaced.

After the recapping, the record-to-reproduce signal path through the electronics seemed pretty good - fairly flat frequency response and low distortion. There were some anomalies in the high-frequency response that depended on level settings, but I will get back to this later. I then started to play the old test tape. I could get moderately flat response to 20KHz at 7.5ips by fiddling around with the high and low frequency reproduce equalization but one channel was definitely worse than the other. It turns out that the transport in the roll-around cart had the 1/4 track heads. Looking carefully at the heads showed that they were pretty worn, and also had worn unevenly. A problem with 1/4 track heads is that the upper track laminations are right at the edge of the tape, and wear differently than the lower track, which is near the middle of the tape.

Looking at the heads in the other transport, which were lightly-used 1/2 track heads, it was clear that a head transplant was in order. Boy - are the head assemblies in the AG-440 so much easier to work on than the 300, 350, and 400-series! It was basically a case of unplugging, unscrewing, moving over the head assembly, then tweaking the tape alignment. The 1/2 track heads do not look like the Ampex type - this is a picture of one:

With the 1/2 track heads, playback is now pretty even between channels. However, there still is this nagging issue of uneven frequency response as the record level controls are changed. This will be covered in the next part.

- John Atwood

Part 3, posted 5/27/01

Thanks Richard and Cary for your comments. Since I design hi-fi and pro-audio equipment, I am pretty familiar with the problem of too-high valued volume controls. I have run tests (on a consulting basis) for some tube-audio companies on their equipment, and always run a frequency response curve at -6dB down. Some equipment have quite severe HF roll-off in this condition!

My procedure for cleaning up equipment with metal panels, like Ampexes, is to first get the dust and dirt off with a moist paper towel (I like the Viva towels). If it is really cruddy, I'll use Formula 409, but with care around painted surfaces. If there are things like masking tape residue, remains of property tags, etc., xylene takes these off without being too aggressive on most plastics and paints. I remove the knobs and other small parts and clean them with 409 and then put them in a Heathkit ultrasonic cleaner. This is good for getting dust off of hard-to-get-at places. It sometimes takes the white lines off, though.

The meters in the AG-440C seem to be different than the earlier AG-440s. They have plastic, not glass, fronts, and are recessed into the front panel. The illumination is by a single 1828 lamp driven through a 150 ohm resistor from the +39V power supply. The bulb is right above the meter inside the front panel.

May 19 - 22, 2001:

While running signals through the reproduce line amps, I noticed that the VU meters did not read exactly right. One channel read about 0.25 dB high and the other read about 0.25 dB low. When restoring old equipment, I have frequently come across "meter drift". Not having the schematics for the AG-440C, I had to trace out the small PC board at the back of the chassis (# 4050564-01) to see what resistors controlled the VU meter level. Since I am only interested in +4dBu levels, I found that the 4.7K (R16) was the easiest to tweak to correct the VU meter readings. Temporarily substituting a pot for R16 resulted in 4.42K for one channel and 4.99K for the other channel to bring the meters to essentially perfect alignment. By the way, the carbon resistors in the circuit that I checked were just about right on in value, well within their tolerance rating. I don't plan to replace any carbon comp. resistors if they are within tolerance, except for special situations like the VU meter calibration. I may end up replacing a few resistors in sensitive front-end circuits if they are noisy, but wholesale replacement is not warranted. The resistors in the Ampexes I am working on seem to be all Allen-Bradley types and spot checks show that they have not drifted at all. Equipment from the late 1950s and early 1960s have the most problems with drifting carbon comp. resistor values.

Next part: the details of the variable frequency vs level settings analysis.

- John Atwood

Part 4, posted 5/28/01

Richard Hess clearly explained the cause of variable frequency response vs level control setting. I've found that in tube equipment that any volume or level control above 100K will almost certainly cause HF roll-off, unless driving a very low capacitance. Because of the lower impedances and pot values involved with solid-state designs, this usually isn't a problem, but add enough capacitive loading, and it will be a problem.

Cary: the reason I didn't put in a VU meter buffer (which is a good thing to do) has to do with the philosophy of the restoration. I actually would like to design new electronics for the AG-440, like Bob Starr did, and in that I will definitely use buffering. However my goal now is to bring the original Ampex design to as good as I can make it without changing its "architecture".

May 19 - 22 (continued):

To analyze the variation in response based on level settings, I decided to stick to the simplest path: Through the record input to the reproduce line output (i.e. with the "Input" button pushed). Thus the signal doesn't have to go through the tape. In the AG-440, this path is basically:

1. Record input, through the 100K "Record Level" pot, then through a 4uF electrolytic to the record amp board.

2. On the record amp board, it goes through a 0.1uF capacitor to an emitter-follower, then through a 3.3uF tantalum to the 10K "Rec Cal" pot.

3. The signal then goes to pin D of the reproduce board where it goes through another 3.3uF tantalum to the direct-coupled line amplifier.

4. The line amp then goes to the chassis where the signal goes through a 250uF electrolytic then through the line output transformer.

5. I ran all tests with the 680 ohm termination resistor switched in and the Audio Precision set to 100K balanced input.

A summary of the effect of various combinations of Record Level and Record Cal is given in this pdf file: The list of the trace colors is in the same order as the comments below it.

What is weird is that there isn't just an HF roll-off, there is some HF peaking! This will be explained in the next part. The roll-off at both 20Hz and 20KHz is over 1 dB, which could make low and high-frequency equalization a bit squirrly.

- John Atwood

Part 5, posted 5/28/01

May 19 - 22, 2001 (continued)

I don't have any extender boards (note: I'd be very willing to buy a set, if available!), so I tack-soldered a few wires onto strategic points on the record and reproduce boards. I have a special low-capacitance, high-impedance cathode-follower buffer that I can connect to the Audio Precision that has less than 10pF loading, and used this for the on-board measurements. Of the several things I found was that the line output capacitor and transformer were not contributing to the weird behavior. The peaking behavior seemed to be happening before the line amp stage. However, when I pulled the record and reproduce boards out on the bench and powered them up, they had very flat frequency responses. I then plugged the boards back in and found that the frequency response at the emitter-follower output was different than the line-amplifier input. On the AG-440(A) schematics I have, these two points are connected through the Output Selector switch. In the "Input" position, there should be a straight connection. What was going on???

I pulled out a spare AG-440C chassis and started tracing the signal connection between the record and reproduce board. What I found tucked in next to the push-button switch assembly was a little network that consisted of a 10K to ground, then a 10mH paralleled with 110pF, then a 10K and a 620pF to ground. The 110pF/10mH network would act as a notch filter at about 150KHz - near the bias frequency. The two 10K resistors to ground would severely load down the 100K Rec Level pot. I first thought that maybe the Ampex engineers were trying to do the trick of turning a linear pot into a log pot by shunting the wiper to ground, but the 100K Rec Level pot is already a log pot. The variable loading (from about 100K to 5K) would explain the variations in low frequency response. The 620pF to ground would explain the HF roll-off. The HF peaking could be caused by the 10mH choke interacting with the 620pF shunt capacitor.

The big question is why is this circuit here? The bias trap is probably to reduce bias leakage in Sel-Sync mode (the sel-sync signal is switched in right before the trap network). There doesn't seem the be the Sel-Sync Bias Trap adjustable trap at the record head input the is shown on the AG-440(A) schematic. Was the 620pF capacitor added to counteract the peaking that the new trap created? All-in-all this is a pretty crappy circuit. However, in the interest of historical accuracy, I will stick with this circuit. In any new designs, I will try very hard to make sure controls don't interact or have other unexpected side-effects.

Next part: reproduce alignment.

[note: I will be travelling for several days, so may not have the next installment until next week. I will still read the email, but will be away from the test bench.]

- John Atwood

Part 6, posted 6/7/01

Reproduce Alignment, part 1 (June 5-6, 2001):

I initially started doing the reproduce alignment with the old 1983 7.5 ips test tape, but once I got the azimuth right, and verified that the reproduce channels were working well, I switched over to the new test tapes. These tapes (MRL 233-423-450-105 for 7.5ips and 243-423-510-101 for 15ips) don't have voice announcements between tones, and are set up to allow an Audio Precision System 1 or System 2 to run frequency response sweeps. It took a little bit of playing around to convert the original 1986 System 1 test (2HD-FREQ) to the system System 2 format, but the conversion was successful, and I could quickly run frequency response curves.

The first thing I noticed was that the reproduce equalization adjustments were extremely touchy. There is a tremendous adjustment range, and it wasn't too easy to make slight corrections. [Note for new electronics design: make adjustments have narrower range and/or use multi-turn cermet trimpots.]

I then started the iterative process of tweaking the equalization. Bringing the supplied 100Hz and 10KHz tones on the test tape to 0 VU (+4dBu) brought the response pretty close to being flat. However it was clear that there were definitely differences between the channels that couldn't be tweaked away. A snapshot of the response is given in: . The Cyan line is channel 1 (left) and the green line is channel 2 (right). There are definitely lumps in the low frequency response for both channels. However, channel 1 is significantly higher in output around 200Hz and 5KHz, compared to channel 1. It also falls more at high frequencies. Touching up the azimuth can alter the HF response slightly between channels, but basically doesn't fix the uneven roll-off.

To check whether the uneven frequency response is due to the heads or the electronics, I powered-down, swapped the reproduce cables, then powered back up again. Everything had to be re-adjusted, since there was about a 1 dB difference in the 1KHz reference level, as well as differences in both the LF and HF settings. After fiddling around, I got the following response curves: . Except for about a 1/4dB difference in the 3-8KHz range, the curves are basically identical, except that the colors are swapped! This means that the frequency response differences are nearly entirely in the heads, not the electronics. Maybe I should change those Nortronics heads...

- John Atwood

Note: I recently have been occupied with a consulting job, plus I will be driving to the east coast and Canada over the next month. I will periodically check my email, but won't be able to work on the AG-440C until I get back in July. Look for more journal entries then.

Part 7, posted 8/9/01

Well, it's been about 2 months since the last journal entry. I've driven from Nevada to Prince Edward Island, Canada, and back, with stops in Connecticut, Maine, Vermont, and Montana. I've completed most of a "high-tech" consulting job, and finally have time to get back to the Ampex AG-440C. The Audio Precision System 2 was also upgraded to a "Cascade" version - this will be helpful in harmonic analysis and digital audio work.

Last time, the 7.5ips response was checked, with some questions about the quality of the Nortronics heads. On Dale Manquen's recommendations, I built a flux loop. Very helpful was Dale's excellent article in the "Handbook for Sound Engineers; The New Audio Cyclopedia". A web search also turned up Dale's very interesting "Tutorial for Tape Recorder Frequency Response Spreadsheet" at

Pictures of my flux loop are at:, and It has 25 turns and is designed to be driven by the Audio Precision at +20dBu. Experiments show that the frequency response is not very dependent on placement. I carefully placed the loop on the reproduce head to give equal response at 1KHz, about -3dB VU per channel.

The playback response at 15ips using MRL tape 243-423-510-101 is shown at This was after tweaking the 15ips response for the flattest response. Note the "head bumps" below 500Hz. Also, channel 2 has a dip around 5KHz, but then rises to 20KHz. The small size of the heads definitely affects the LF response. The unequal HF response between channels is puzzling.

The response from the flux loop is at This has not been corrected for playback equalization, so it rises above 1KHz. The peaks at around 22KHz must be due to resonance of the head inductance and stray capacitance. This definitely sets an upper limit to playback response. The droop below 200Hz is likely the LF equalization.

The interesting thing is that except for slight differences in the amount of equalization, the flux-loop response for both channels are nearly identical. This points the blame for the unequal HF response to the heads or head alignment.

- John Atwood

P.S. Previous AG-440C Journal entries are collected in This will be kept up-to-date as new entries are added.

Part 7.1, posted 8/9/01

This journal entry is to correct and clarify part 7. Thanks for everyone's comments.

First, Jay, you did find a mistake! My comments on the bottom of the flux-loop frequency response are backwards: swap green and red. I have updated the pdf file to correct this. I was aiming to have all graphs show channel 1 be green and channel 2 be red.

Regarding the "similarity" of the two curves in the flux-loop response, my point was that they are nearly the same shape, with the differences likely due to different equalization settings, which were an attempt to compensate for the weird HF response. Keep in mind that I aligned the reproduce response as best as possible before doing the flux-loop tests.

Thanks for the pointer to . I haven't tried every thing you listed, but I did try running the tape backwards. The composite frequency response curves are at The reverse response is nearly identical to the forward response, although there is about a 1/2dB difference at 20KHz. I don't think the tape is getting magnetized, because the HF responses have been pretty consistent. The test tape was unused (virgin) until last night when I started the 15ips tests.

Bo: Channel 2 is the lower of the two tracks in the head stack. I will look at the tape alignment some more, but am beginning to think that the Nortronic heads are strange. I think I will try to get new heads. For now, though, I will continue alignment, just noting the strange response of channel 2.

- John Atwood

Part 8, posted 8/10/01

I did some more research on the funky Nortronics head. I unplugged the head right at the head assembly and did some measurements on the head. The DC resistance was measured by a Fluke 87 DVM and the inductance and Q measured by a General Radio 1650-A impedance bridge. I got the following readings:

top track bottom track
DC resistance 4.5 ohm 4.7 ohm
Inductance 4.3 mH 4.0 mH
Q 4.5 4.0

The bottom track definitely is different. I don't know what the usual tolerances are, but these definitely are not a matched pair!

I then ran a frequency response sweep with the flux loop driving the head, and the head output going directly to the Audio Precision input. The results are at . Except at low frequencies, the response is very close. The LF differences may be due to the exact placement of the flux loop, which is hard to control.

After all the testing of the head, I figured it was time to give it a good demagnetizing. I normally demagnetize the whole tape path before running test tapes or after doing things that might result in transients (like pulling out a powered-up board!!). While I had the head leads connected to the analyzer and scope, I thought I would see how good my demagnetizer was. I have been using an old Olsen electronics demagnetizer made in Japan. Monitoring the waveform on the scope showed that the waveform never even got close to distorting, showing that the head's core was not saturating! Maybe I have magnetized my test tapes 8-(. I dug around in some boxes an finally found an old Ampex demagnetizer unused in its original box (part number G-704-00). This one gave a much stronger field, and could cause noticeable distortion in the output waveform. I gave everything in the tape path a good demagnetizing.

Next time: rerunning the test tapes.

- John Atwood

Part 9, posted 8/13/01

Brian Roth sent the schematics and alignment info for the AG-440C and Dale Manquen prodded me into getting the NAB equalization entered in as an "equalization file" for the Audio Precision. The AG-440C alignment specifies the use of a flux loop, and has the procedure for adjusting the head resonance control. I made up an Excel spreadsheet that calculates the correct inverse equalization for the 50 Hz LF corner and the 3180 Hz HF corner for 75 points from 20Hz to 50KHz. If anyone is interested in this, let me know, and I can send it to them.

After running the Ampex alignment procedure using the flux loop, I got the following *pre-equalized* curves:, and for 7.5 and 15 ips, respectively. Ideally these curves should be identical, since the NAB equalization is identical for both 7.5 and 15 ips. Note that above 20KHz and below 30Hz the curves are essentially identical. The rise above 7KHz is the head resonance, and is normal, according to the Ampex information. The head resonance control moves the peak of the resonance a maximum of about + or - 7KHz. On the whole these curves look good.

Getting ready to re-run the test tapes, I thought I would double-check the tape alignment. The reproduce head was very slightly tilted, so I started fiddling with the allen screws and azimuth hex nut. One thing I noticed was that I had to do the alignment with the tape running, since the resting position of the tape on the head moves slightly, compared to the static position. It was then that I noticed what is likely the cause of my problems. There is a very narrow lip on the top of the top track. The head has slots above and below the tape track. Apparently, in previous usage, the tape was slightly mis-aligned and there was a little piece of unworn head track at the top edge of the top track. If I centered the tape on the head, the tape is pushed away from the head at the top of the top track. This is visible while the tape is moving. If I moved the tape low enough not to be disrupted, the tape was clearly not centered. Since I had the tape centered prior to my alignment, apparently I had compensated for the unequal coverage on one track by using the HF equalization. However, it could not compensate this kind of defect!

At this point, it doesn't make sense to go forward without either getting the heads relapped or getting new heads. I have a request into JRF Magnetics for these options. The AG-440C journal will take a break until I get the new heads in.

Everyone's help so far has been great - thank you!

- John Atwood

Part 9.1, posted 8/17/01

This is just a follow-up to Part 9. I took the heads out. The Reproduce head is shown in and The "P" is my own marking. Otherwise there are no part numbers marks. The Record head is similar. It was really hard to photograph the offending lip at the top of the top track, but is a try. The arrow points to the lip.

I got a good quote from John French (JRF Magnetics) on relapping and new heads. I think I will get the current heads re-lapped, but may also get some new Flux Magnetic heads. I'll keep everyone posted on the progress.

Regarding Jay's comment, yes, I meant that the curves were essentially the same between 30Hz and 20KHz. (I've always had a touch of dyslexia and tend to accidently invert things. It made me a lousy digital logic designer until I discovered logic simulators!).

The AG-440C alignment instructions say to set the 7.5ips response to the following levels:
500Hz +0.2 dB
5KHz +5.4 dB
15KHz +15 dB

The 500Hz and 5KHz values are right about on the pre-equalization curve. For 15KHz, the pre-equalization curve calls for +13.75 dB. This is 1.25dB below what the instructions specify, so I adjusted the level to about 1.25 dB above the pre-equalized 0 dB floor using the resonance control. (Actually, now that I look back on my measurements, I think I used +16 instead of +15 dB - I will double check this later.) I found that the resonance control seem to affect the frequency of the peak more than it did the Q.

- John Atwood