Akai/Roberts Preamp Mod - Roberts 720A and Akai M7 (and others)

Last Updated: Tuesday, 27 February 2024

Many people have modified tube-based reel to reel tape recorders from the 50's and 60's into microphone preamps and guitar amps. (You can see a cameo by one of these Akai/Roberts decks in Apocalypse Now -- not sure which model that is.) Rod Childers at Beyond Sanity Productions provided exceptional instructions (his website appears to be dead now -- this site now hosts most/all of the info/files that were on the Beyond Sanity site). I did this for a Roberts 720A and an Akai M7. I made this page for anyone trying the mods on the same models, and to provide information on the subject in general, since the M7 is commonly modded, and the 720A is basically identical to the other commonly-modded models.

Note that I am a rank beginner at all this, so take it with a big grain of salt.

If you aren't going to use the neon lamps that come with these amps, would you kindly contact me before getting rid of them? Thanks!

The schematics, user manual (including a parts lists and annotated images of the innards), recent pictures of some mods, and instructions for hum control adjustment can be downloaded here. That includes an annotated version of Rod's step-by-step instructions, in which I corrected/clarified a lot of things that were confusing to me. If you just want to see the latest schematic, here is a link.

BACKGROUND

WARNING: This is very dangerous. Tube amps can easily kill you (250V-350V+ are exposed in these). Take it seriously, and don't try these modifications unless you are qualified. See also these safety notes in the build notes.

I did a normal "2-in-1" mod for both. I added a switch for the power tube section, used existing switches for the backlights in the meters, and for the 720A added an impedance selector switch for the mic input transformer (all very simple adjustments to the standard mod). I used Cinemag CMMI-10C input transformers on the 720A, and Jensen JT-115K-E's for the M7. I modified the original case to house the amps, which works great, but does take some time.

The 720A works fine for these mods: same tubes, same circuit, etc. There were some anomalies that I don't understand, but I suspect those can be attributed to my lack of experience, not the amps themselves.

Note: It is normal with these mods for each of the stages to be polarity reversed: meaning, an audio signal sent in will be "flipped" in polarity on the way out. Though basically harmless, this can be inconvenient if you're mixing the signal with other, non-flipped mics on the same source. You can flip the polarity of the XLR input (see details in the annotated instructions, linked just above), but the 1/4" jacks (including the 1/4" input on the EF86 stage, if you have one) will still be reversed.

Note also that the output impedance of at least the EF86 stage might not be what you expect... Going into a high-impedance input like some on audio interfaces (e.g. the 1/4" input of the Mic 1/2 jacks on a Presonus Firebox, which have 1M-Ohm input impedances for guitars, etc.) can result in loss of high frequency and high sensitivity to cable length (longer cables leading to more loss.) If you have any insight into this issue (e.g. what the output impedance of these units is, exactly), please contact me.

On the subject of loudness: used as a guitar amp, going into a 1x10 or 1x12 8-Ohm speaker cabinet (those are the sizes i've tried), these little 3-5 watt power amps are able to be respectably loud.

MODELS

Working models:

Roberts 770, Roberts 770A, Roberts 770X, Roberts 720 (very likely, see below), Roberts 720A, Akai M7, and Akai M8 are known to work. I did see one picture of a weird-looking 770, so I'm not sure those can always be guaranteed.

The Roberts 990 looks very likely to work (one person described it as "the same as M8").

The Akai M-7SP and Akai M-7SE look very likely to work, but that's unconfirmed. The SE apparently was the first with "crossfield" heads, and the SP might have just referred to external speakers that came with the unit, not sure. The Akai 707 seems very similar... not sure about it though.

The Roberts 997 is often conflated with the Akai M7. Looks the same.

Note: there is no reason I'm aware of to favor a particular model over another in terms of these mods -- they will all be functionally identical. There may be some slight differences in layout, slight differences in tube sub-type, etc, but there is no reason to choose one over the other besides the aesthetics of the dials, faceplate, etc. As far as I know, the differences between these models were all in the tape transport and other tape-related features that will be removed by the mods. I mention this because some people search only for an M8, thinking that it's "the best" when I'm not aware of any reason to do so. It's possible that there is some advantage to choosing the latest model: for example, perhaps the later date of manufacture means components are less likely to have failed. But the opposite could also be true if the manufacturing cut corners over time. Until someone reports in about such differences, I recommend you cast a wide net and find any unit you can if that is in good condition for a good price.

More info:

Roberts 770A, 770X and Akai M8 are all the same (says helpful internet person Adam: the only difference is the faceplate; the 770A came out in '63-64 and the 770X in '66-67). The 720A is the same as those but without some tape recorder features. The "770X SS" is a solid state version (no tubes, obviously not appropriate for these mods).

Roberts 770 and Akai M7 are the same. I think they are cheaper versions of the 770A/770X/M8 (but that should be irrelevant for the mods). The 720 is apparently some kind of cheaper version of those (akin to the 720A-to-770X/M8 relationship) but I'm not 100% sure that a 720 would work for these mods. I assume it would. Let me know if you know.

The Roberts 90C (90-C) might be an earlier, mono version... similar jacks and knobs, one had 12ad7 and ECC83; another was reported to have: "6AU6, 6AU6, 6X4, 6BQ5, 6AR5, 12AD7" (the inside pictures looked pretty different in some ways, pretty similar in others. Might be different enough to require some expertise to make it work. Not sure what the extra 6AU6 was for - it seemed to be mounted on some separate piece of the case?)

Models that will not work:

(It's been a while since I made this list, but I think these are all Roberts model numbers except where noted.)

The "definitely wrong, or you should assume it won't work" list includes: 5050X, 420XD, 333X, 450, 450A, 1720, 1721, 1725W-III, 80, 525, 530, 100, 95, 333X, 400X, 440, 5050XD, 420X, 650XD, 610X, 144, 1700, 1719, 1740X, 6000, 450, 1650, 1670, 770X SS, 771X, 778X, 800X, 808, Akai M4, Akai M5, Akai M9, Akai X-100D.

The "has or might have tubes but probably not workable with the standard mod" list includes: 1040 (has tubes), 1055 (2 X 12AT7, 2 X 6BM8, 6CA4), 1057 (tubes), 1600 (tubes), 1620 (tubes), 1630 (tubes, made in 1966 at least), 1640 (made in 1965 at least, has 6bm8, 12ax7, 12at7), 1725-8L III, 1725

Possibly-useful links with images of various models:

https://www.reelprosoundguys.com/roberts.html
https://reel2reeltexas.com/vinListA.html

NOTES

There are lots of little notes and suggestions and so forth in the annotated instructions (linked above). Make sure to read those thoroughly.

Mistakes/anomalies in the schematics:

the left amp schematic of the original machine included in the mods ZIP file ("Akai M7 Left Schematic.jpg"), has a couple anomalies: the lead from the right side of R20 looks as if it connects to C19 - it does not, AFAIK. Look at that part of the schematic from the 2_Preamps_In_One schematic Rod provides (or in the Akai service manual schematics): it actually dodges around C19 and connects to TR2. Also, this schematic, and others, presumably, don't label L1, which is the oscillator coil connected to C21, C22, and C23. And "TR3" in that image is also known as "L2". For my 720A, J2/"PHONO RADIO" was labeled "Line" and J3/"HEAD OUTPUT" was labeled "PHONO PLAYBACK". For the M7, the J5/"PREAMP OUTPUT" was labeled "PICK UP".
C16 is listed incorrectly in the parts list (and maybe some schematics?) as "25uF/25V". In my amps, and in the main original schematic, it is actually 20uF/300V.
The 720A and M7 hum control connections were a little different. The 720A center wiper of the 12AX7 control was connected to a point common to R17, C15, and pin 3 of the 6BQ5/EL84. My notes are unclear as to how the other connections went.
there are some diagrammatic anomalies regarding C21 and C24 -- see the annotated documentation (linked above) for more info.
In one of the pictures on Rod's site (DSC_7673_12AX7_IN1.JPG) the tip is mislabeled "Tip of 12AX7 out"; it should read "Tip of 12AX7 in".
on the 720A there were two 47uF caps in series from one of the wires off the switch between the two amps; didn't see them on any schematic or in the instructions -- i removed them. They were not present in the M7.
one of the meters on the 720A has a capacitor between the terminals. I just left it. Not sure what it's about, but the meters both work fine (and the same, AFAIK).

Confusing things in the schematics:

the connection of the 12AX7 out to the 6BQ5 (jack J5) is confusing for some people (and was for me.) Rod has some info (and pictures) at the end of his post on his site (but not in the instructions). Basically, the schematic is a little misleading (if technically accurate): the yellow wire from the 12AX7 jack is switched: when a plug is inserted, this wire connects through the switch to ground. The yellow wire connects to pin 1 of the 6BQ5 (not pin 2 as is implied by the schematic). On 6BQ5s, however, it's very common for pins 1 and 2 to be tied inside the tube. Pin 1 is used as "the mute" for the tube: when a plug is inserted in the 12AX7 output, it mutes the power tube by tying the control grid to ground. The green wire from the 12AX7 output connects to C30. The schematic makes it look like the yellow wire joins with the lead from C11 and goes to the tube, but you will see the yellow wire going to pin 1 all by itself (but it is electrically joined in the tube, so the schematic is accurate, afaik). In summary: tip goes to C30, sleeve goes to ground (and when cable is inserted, 6BQ5-pin-1/2 and C11 also go to ground.)
Modder Dale reported an M7 that had no C30 at all. It was otherwise apparently normal. (My M7 did have C30 as expected.)
The instructions say that pin 1 is the input for the 6BQ5, but that the input is spec'ed as pin 2 (afaik), but since they are tied together it's the same thing (spec sheets will show pin 1 as "I.C." - i.e. internally connected, proprietary to manufacturer, etc, but I take it that the pin 1/2 joining in 6BQ5's is near-universal.)
L2 is shown on the schematic as "TR3"

Build Notes:

In case it's useful, here is a video of the interior of my AKAI M7, right channel. The reason for this video was to show wire routing and such for a fellow builder; I make no claims about the prettiness of this build. :-)
I used combo XLR/1/4" jacks for the EF86 input - the mounting holes required a couple extra holes be drilled, and it's awkward to fit the nuts on the back side against the body of the jack, but it worked out. The tip of the 1/4" goes to the same junction with the pot as does the secondary from the transformer (not ground side), and the shield I tied to ground (when I didn't tie the shield to ground, got tons of buzz/hum). The schematic of Rod's "Boris Model" shows the wiring of the tip, but not the shield to ground: I believe that the jack he used is grounded to the chassis, but you'd have to ask him.
I used a hole saw to get through the face plate. Worked fine. (i.e. you don't need to buy a stepper bit if you don't have one). Clamping a wood block to guide the hole saw helps.
for the record, the set screws for the original knobs are 10mm long, 4mm diameter, with a 0.75mm thread pitch. Basically unobtainable (i tried, hard) without having them custom made, so be very careful when unscrewing them, as Rod recommends. The set screws for the concentric knobs are the same diameter/pitch, so if you're not using the concentric knobs (and you won't be if you do these mods) you can use them as replacements if need be.
biasing of these amps is not needed when putting a new 6BQ5 in. The technical reasons why are beyond me, but so sayeth Rod.
Rod had a post in his DIY section about adjusting the hum trim controls for the preamps -- that link now goes to an Internet Archive page, and the hum adjustment info is included in the .zip file at the top of this page.
On the 720A, I used Cinemag CMMI-10C's for the input transformers (selectable input impedance) and a 100k variable resistor for the pot (instead of the 500k in Rod's mod instructions). On the M7 I used Jensen JT-115K-E's and a 250k pot with a 383kohm resistor in parallel to simulate a 150k variable resistor.
For each set of pres, i got a Rolls phantom power box off eBay (PB223) to provide phantom power.
as others have done, my main power switch is a 3PDT on-off-on that switches the 6BQ5 power tube in and out of the circuit: center is off, up is all on, down is pre-amps only on. It switches the mains power to the transformer, B+ to the 6BQ5, and one leg of the heater current to the 6BQ5. (Re: B+, I'm not looking in the amp now, so don't quote me on this, but I assume I maintained the connection between C14 and C19, and just interrupted the leg to B+ on the 6BQ5.) Besides using less power and making less heat when you don't need the power tubes, it gives you the option of engaging the power tubes when you want "power supply sag" (which is a bit like compression, to my ears), and the ability to see the meters dance.
I repurposed the monitor push button on the left channel and the old mains power switch on the right, as switches for the backlights of the meters.
I used a 2A fuse in a case that houses both amps. Ripped the vinyl (?) covering off the original case, cut it down, reinforced it with some of the old hardware, and sanded off the glue.
I shocked myself once. Frightening. Take it seriously.
Beginners take note: everything you read about careful routing of AC legs and shielding and wires in general is true. These amps were great radio receivers when I first powered them up, and now they are totally quiet simply from finagling the routing. Of particular concern: there is a pair of twisted wires feeding the B+ to one of the center tubes (can't recall which off hand - 6BQ5?) - it can cause an overwhelming hum if it's near anything sensitive and needed to be routed carefully. I ran the twisted pair from the 12AX7 output in front of the tubes (you can see it through the grills from the front) and down to it's connection point on the board keeping as far away from other stuff as possible.
in terms of safety, it has been said that the large capacitors are the only ones needing draining and monitoring to avoid latent charges shocking you even when the unit is off and unplugged. In practice, I saw the capacitors bleed voltage on their own to a safe level after 10 or 20 seconds, so I assumed that there is a "bleeder resistor" functioning somewhere, but to be safe I still used a lightbulb-style discharging method before operating on the amps.
Another safety concern: the old resistors seem to have a very thin (or worn?) coating. The resistor pictured below (R20) had ~350 VDC present on the coated rim. My multimeter showed continuity between that part and the bare metal leg. I assume the other resistors are similar. I covered the ends of R20 with electrical tape just because it sticks out in a way that feels a little unsafe, given the voltages present. Not sure if other components in the amp are the same way, so: be careful.

Further improvements/ideas:

If you intend to chain the two stages together, I recommend adding an additional pot between the EF86 and 12AX7 stages to act as a gain control for the 12AX7 stage: it allows for significantly less hiss. This is described in more detail in the annotated instructions in the section on the 12AX7 stage. Here is a picture of the additional knobs added to my 720A amps. You can also see the thread on gearslutz where people helped me figure this out.
Rod now recommends a 0.22 uF capacitor off the EF86 to replace the .15 shown in his instructions (this is noted in my annotated instructions linked above)
On the 720A I had a crazy buzz problem whenever the power amp was switched in and a plug was inserted into the 12AX7 output and the gain went up. The meters would also die completely. Never figured it out, but when I defeated the switch at the 12AX7 output with some electrical tape (thus preventing a plug from grounding pin 1 of the 6BQ5 and muting it), this problem went away completely. Since i don't need the muting feature, and i like the idea of being able to take a DI from the 12AX7 and still use the power amp output, this suits me fine. I have no idea what the implications of this change are, but things seem to be working fine.
Rather than hard-wiring the polarity of the XLR input to the EF86 stage, one could wire up a DPDT polarity switch.
A helpful person on the internet suggested a few additions to improve the amps (see that link for more info): add a 10k to 20k resistor from C30 (negative side) to ground, and add a 1k to 2k (i did 1.8k) from C30 (negative side) to J5-tip ("1.8 k is fine. With a 1.8 k output impedance you can still drive up to a 2000pF load without serious HF loss. Even a 20ft. or 30ft. cable is OK.") My ignorant estimations are that 1/4-watt-rated resistors are appropriate here, but I'm not 100% sure. More info: "The 20k to ground will provide a DC path to properly charge the cap when the amp powers up, so if you plug that tap into an input it won't create a gigantic transient pulse. 1k will will give enough isolation to the output and will still give a reasonable output impedance." Addition of these resistors made no negative change in the sound that I noticed, and presumably will help in the ways described.
Rod suggests in his instructions that increasing the capacitance of C2, C10, C15, and C30 may decrease the "compression" effect you might hear in the amps.
There is a neon indicator lamp in some of these amps (originally used as the record indicator). (If you have any from amps you are modding and you don't want them, please contact me!) They do not appear in the schematics (though they are listed in the parts list as "PL2 Neon tube bracket neon"). It's not 100% clear, but they seem to tap in (through a 300k resistor hidden inside a braided insulator sleeve -- also not on the schematic) to a line that operates at ~65 kHz and 100-200VAC. At any rate, to use them as power indicators (see image here), I connected them to the 117VAC tap on the main transformer, in series with 400k resistance to ground. You can lower the resistance value to get the lamps to glow brighter, but I didn't see much of any improvement between 300k and 400k, so I kept it at 400k to maximize longevity of the bulb. (If you connect these lamps with too-little resistance they will destroy themselves.) If your mains supply is 240VAC, you can double the resistance, but one person told me you may want to go even higher than that, since the strike voltage for the lamp doesn't change and there is now a greater difference between the strike voltage and the mains current. Let me know if you experiment with these bulbs on 240VAC. Dirty details: on the component strip in at least one unit, the components at the end went C15, C24, C21 (C21 is a brown "AK"-like package, which will be familiar if you've read through the annotated instructions). The last couple are not pictured in some of the service manual images. One leg of the lamp goes to ground, and the other goes through the 300k resistor to C21 (on the side not connecting to the tube.) This connection also leads to the top-right slot of PG1 (in one of the schematics in the service manual -- I suspect the other is in error?) which connects to the top-right of JC1, which goes through the trimmer pot C27 to the note "65kc 100~200V". If you look at service manual PDF page 34, it shows C21 in the "left" image -- they seem to have moved it to the circuit board in some units. Thanks to modders Tom and Hanno for the assistance in figuring this all out.

PARTS LIST

I don't necessarily recommend these sources, they just happen to be where I got my stuff. This is for the 720A mod that I did.

This assumes that you're wiring the 12AX7 input for guitars (see instructions .rtf file).

Make sure to see the further improvements section before buying your parts; especially the note about the extra pot between the two stages.

Two Cinemag CMMI-10C mic input transformers: https://cinemag.biz/mic_input/mic_input.php
Shielded wire: I got 10 feet of the RG174 cable in the "Wire/Cable" section of https://www.hoffmanamps.com
Knobs: You'll need two knobs that fit a 1/4" shaft. I got some original Roberts knobs on eBay.
switch for hi/low impedance taps on mic input transformer:
https://www.mouser.com/ProductDetail/Mountain-Switch/108-2MS1T2B3M1QE-EVX/?qs=sGAEpiMZZMvudeGI7i40XPFFoGoDDcZUdmcMmVM4VMg%3d
2 combo XLR-1/4" jacks for EF86 mic and hi-z in:
https://www.mouser.com/ProductDetail/Amphenol-Audio/ACJC5S-BULK/?qs=sGAEpiMZZMv0W4pxf2HiV9DEPBIg8Cuf9AveNx6bkgY%3d
Switch for cutting 6BQ5 power tube - 3PDT on-off-on (this one is tiny, but the more appropriate sizes were much more costly):
https://www.mouser.com/ProductDetail/Mountain-Switch/108-0006-EVX/?qs=sGAEpiMZZMvudeGI7i40XPrKyNLVhQVgXbJBwkVJzmQ%3d
1 IEC socket with 5x20mm fuse holder:
https://www.mouser.com/ProductDetail/Schurter/62002300/?qs=%2f0Eb3QIYuIYjaCVWTqCP1w%3d%3d
A 5x20mm 2A fuse:
https://www.mouser.com/ProductDetail/Cooper-Bussmann/BK-GMC-2-R/?qs=sGAEpiMZZMseCiJT91fwIlyoGrb8FM5n8Fa%252biwY8nYI%3d
3 of these terminals for the power leads going to the IEC jack:
https://www.mouser.com/ProductDetail/Molex/94030-3891/?qs=sGAEpiMZZMvz8LftK4jerskvOOK94mtbeyubf41up1s%3d
3 of these M-M-F adapters to attach to the IEC jack to provide connection points for the power leads:
https://www.mouser.com/ProductDetail/TE-Connectivity/726308-1/?qs=qCxwlXJ4fnzL4cMpJYCISA%3d%3d
2 100K Ohm audio taper pots (for Cinemag input transformers - must match your input transformer appropriately); 20%, serrated shafts:
https://www.mouser.com/ProductDetail/BI-Technologies/P231-QC20AR100K/?qs=sGAEpiMZZMuarXojuOTsc5GHFIFyOjUWf9sTQ1wzifI%3d
2 .22 uf orange drop caps (400V):
https://www.mouser.com/ProductDetail/Vishay/715P22454MD3/?qs=b4iwvj6kQLoRJnN5Xm9nnA%3d%3d
2 68k resistors (for 12AX7 input in "guitar mode" (see Rod's instructions):
https://www.mouser.com/ProductDetail/Xicon/273-68K-RC/?qs=sGAEpiMZZMtlubZbdhIBIOP1vYJrXbgiR3NluvU%2fRSk%3d
2 1M resistors (for same 12AX7 input):
https://www.mouser.com/ProductDetail/Vishay-Dale/CCF601M00FKE36/?qs=sGAEpiMZZMu61qfTUdNhGyW2LHYvMinE8yS3HPU31OA%3d
I used 50 mil solder 63/37-RA - 0.050 inches, 1.27mm, about 1/16th":
https://www.mouser.com/ProductDetail/Kester/24-6337-0053/?qs=m5JL4TdtMhzw0hxlVK4REg%3d%3d

TUBES

These units use the following tubes:

first stage: EF86, AKA 6267 (6BK8 is related)
second stage: 12AD7 (though often they seem to have a 12AX7 in them, possibly as a replacement. 12AD7 is a "non-microphonic/low hum" 12AX7. I think it has the same gain. It's pretty hard to find them.)
6x4 - full-wave rectifier (AFAIK: 6x4 = 6Z31 = EZ90 = CV493 = U78 = 80BK = KX-80BK = V2M70)
6BQ5, AKA EL84 - the power output tube
6AR5 is unused after the mod

I replaced a few tubes with NOS tubes from the internet. I also replaced all the tubes in one of the amps with new (not NOS) tubes. Tubes certainly sound a little different from each other. The replacements I made didn't, however, make an appreciable difference on background noise of the amp. I also got a 12AU7 and 12AV7 to experiment with lower-gain tubes when chaining the EF86 stage into the 12AX7 stage (see below). They definitely bring the gain levels a little closer to sanity, but the 12AX7 was still entirely usable in that configuration.

When I chain the two input stages together (EF86 into 12AX7) sometimes I get crazy oscillation feedback when the power amp section is switched in. Rod says: "When you chain the EF86 into the 12ax7 it is a ton of gain, any little issue with a tube or the circuit and it will start to oscillate. When I do this I usually use a lower gain tube, 12AY7, or I have also used some 12DW7s (The first triode is much lower gain than the 2nd)". I realized that I can chain from the EF86 of one amp into the 12AX7 of the other, and a vice versa, and there's no oscillation. And an update: when I use the units as guitar amps, the levels are all low enough that this isn't an issue, and I can sometimes chain within one unit, but it depends on the unit and tube tubes. I don't like to crank it up loud, though, so I'm not sure what would happen if everything was maxed out.

I tested my EF86s and 12AX7/12AD7's with the help of generous folks at the awesome local radio museum. Some tubes were weak. I think it's a good idea to check them out if you possibly can. Alternately, you can buy one set of newly-manufactured tubes and using those when you are first testing your modded amps, just to rule out tubes as a source of issues. Then you can swap the old ones back in, if desired, and you have a set of spares on-hand.

The tubes that fit use the "standard" tube sockets for amps: Noval/B9A socket, aka "9 pin miniature". However, the EF86 and 12AX7 sockets have the bottom part of an EMF shield that they slot in to (none of the amps I saw had the top part). This shield is a little narrow for modern tubes: it's approximately 27/32 inches wide (21.43mm), and the old tubes I've seen are around 26/32 inches (20.64mm). New tubes seem to be around 27/32 inches and won't fit into the shield. My solution was to take a pair of tin snips and make a few vertical cuts through the shield. Then they can gently be bent slightly outward and the tube fits in fine. Do not drill the rivets out: the whole socket is attached and it will come loose with the shield.

CAPACITORS

I don't have the expertise to make claims regarding capacitors, but a helpful person on the internet guided me through it.

I replaced all the major capacitors in one of the amps. I did it in four stages, replacing them a couple at a time, and recording samples of the amp noise output along the way. It made no noticeable difference on the amount of noise or hum in the amp. Based on that sole data point, I'd say that capacitor replacement may not be necessary if the capacitors seem to be in good shape and you don't have extraordinary noise. For me, routing of wires in the amp was very important, and capacitor replacement apparently useless. See the tubes section discussion of tube replacement.

It cost me about $60 to replace all the capacitors in a single amp (i.e. just the right channel) with general quality, or slightly better quality, capacitors. The big ones are expensive.

Replacing capacitors with others rated at higher voltage is, as far as my limited knowledge goes, fine, within reason, and recommended for the power filter capacitors by my sources, who suggest that mains voltage that fluctuate from the ideal could result in significantly higher voltages in the amp.

Listed roughly in order of priority for replacement (replace the first in the list first):

C25 and C26: these are power filter caps. They are in a "dual cap" housing -- a single shell with a common ground terminal. I replaced them with two individual capacitors because dual caps are hard to find. Per suggestion, I increased C25 to 40uF (you could also use another dual cap 20/20 and add an extra 20 in parallel with C25 to get 40). Do not increase the capacitance of C26, I'm told. I zip-tied them to a nearby part of the chassis which conveniently protruded nearby. Consider that they are carrying high voltages, and shield accordingly.

C16: the "screen bypass". A 350V rating was recommended to me. This is listed incorrectly in the service manual parts list (and maybe in some schematics?) as 25uF/25V. It is actually, at least in my amps, and in the main original schematic, 20uF/300V.

C7: "input stage bypass"

C2,C10,C15,C30: "cathode bypass caps" (not sure if C30 is also a cathode bypass). Rod suggests that larger capacitance values here might lower the "compression" effect in the amps. I did not increase their values (didn't notice his suggestion until afterwards).

C9,C11: "coupling" capacitors - "use the best you can afford here". I used polypropylene 'orange drop' caps. I think 400V would be OK (?), but i got 600V's just in case.

These are the capacitors I purchased:

C25: https://www.mouser.com/ProductDetail/Vishay-Sprague/TVA1712/?qs=sGAEpiMZZMsh%252b1woXyUXj19oZiSK8UqOP56iiRHgBdk%3d
C26: https://www.mouser.com/ProductDetail/Vishay-Sprague/TVA1709/?qs=sGAEpiMZZMtZ1n0r9vR22UzXQwAPJeqEDrBn%2fWOSncU%3d
C16,C7: https://www.mouser.com/ProductDetail/Vishay-Sprague/TVA1608-E3/?qs=sGAEpiMZZMtZ1n0r9vR22WtDl%2f7UhxBOSu0A3RZYCwY%3d
C2,C10,C15,C30: https://www.mouser.com/ProductDetail/Vishay-Sprague/TVA1306/?qs=sGAEpiMZZMtZ1n0r9vR22UzXQwAPJeqECXFITVQbZQM%3d
C9: this was out of stock: https://www.mouser.com/ProductDetail/Panasonic/ECW-F6203JL/?qs=sGAEpiMZZMv1cc3ydrPrF0%252bjlB8SXIRuZtM9xfHCQrk%3d, so instead i got it here: https://www.triodestore.com/sprag71ordro.html
C11: https://www.mouser.com/ProductDetail/Vishay-Sprague/715P10356KD3/?qs=sGAEpiMZZMsh%252b1woXyUXj6uDrZiYrjEyzKBh2zGVi08%3d

LINKS

Here is the list of all interesting links I found on the internet about these mods, roughly in order of usefulness. There are more, and I haven't updated this list since I first created it. You could search YouTube as well.

These beyondsanityproductions.com links are broken, so I've provided them as links to the wayback machine archive with the last good capture: