Optimizing old tube guitar amp's routing of analog 0V

[Matador]

I've dealt with a dozen or so requests from people who own older tube guitar amps, who are unhappy with the level of noise and hum that they have to deal with, especially for recording in studios. I see a lot of questionable layout practices and wanted to bounce my solution off of the group to see if I was missing anything.

Taking a mid-60's Fender Deluxe Reverb for example: Leo made extensive use of the chassis as a global analog 0V node. The biggest culprits here are a) returning the rectifier filter cap through a portion of the chassis, and b) return the power tubes cathodes through the chassis. There are also myriad connections spread throughout the preamp. In later years, a brass plate was added behind the pots and 0V connections were not only bonded there, but also to the chassis along its length with large solder blobs.

This seems sub-optimal, although I understand it was to ease the burden of assembly (e.g. lower cost), and minimize ground impedance (likely a side benefit after cost).

First attachment is an AB763 layout annotated with orange circles for every implicit/explicit bond of analog 0V to the chassis. There is a large loop area between the first rectifier cap, and also the cathode return currents (perhaps less of an issue, due to the fact that the currents in each side are balanced). Second attachment is my optimization:

1) Add a 12 gauge copper bus bar (stolen from 12-2 Romex) between the pots and the eyelet board, held off from the chassis a bit with insulated standoffs.
2) Reroute the analog 0V and treat it as (mostly) a differential signal. The chassis and the circuit are completely isolated from each other, with the exception of a single connection (at the vibrato input jack).
3) All pot cases are bonded to the chassis via the mounting nut and lock washer - leads on pots that were bent over and soldered to the pot case are lifted and run over to the bus bar with short wires that run along the chassis
4) The rectifier loop is limited to the end of the bus bar.
5) All cathodes are returned to the bus bar.
6) The two input jacks are changed to shielded wire, and the shields are tied to the chassis at the jack end (to become extensions of the chassis).
7) The speaker jacks are changed to insulated jacks, and only connect to the output transformer secondary. To make the feedback work, a separate wire is taken from the output jack's '0V' and connects to the same place as the 47 ohm feedback return 0V point (a poor-mans remote ground sensing).

This is still a compromise, because the first rectifier cap is still up in the amp's doghouse, so the "send" and "return" wires between the power transformer secondary, rectifier center tap, and the main filter cap, are free to radiate rectified noise inside the chassis. It might be worth it to relocate that main filter cap to the inside next to the rectifier to minimize the worst loop area.

The only thing that is weird is the reverb tank: the RCA jacks' secondary are all bonded to the chassis, however the send is low-ish impedance, so I don't know how much this matters in practice. Inside the reverb tank, the input/output jacks are also tied to the reverb chassis. I don't see a way of breaking this without moving away from RCA / unbalanced connections for the send / return (so the reverb transformer's secondary can be sent differentially to and from the reverb tank, and the chassis/shields are connected separately and become an extension of the main chassis.

Is there anything I'm missing here? Or is this tilting at windmills?

 

[dmp]

I have a '64 deluxe reverb and I don't think it needs a mod. But it would be interesting to do an A/B test to see if there is a change. In a tube amp you have a hiss floor from the tubes that might hide any difference from the ground layout. Ultimately, the right answer is when the amp is quietest even if it has multiple connections to chassis.

From the section on bus grounding in Merlin Blencowe's book on designing tube preamps:
"One logical ground scheme is the bus ground. This involves routing a single, heavy gauge wire, called the bus wire or bus bar, through the chassis. The path of the bus wire should follow the natural path of the amp circuit from the reservoir circuit from the reservoir capacitor, to power amp, to preamp, to input stage, and all ground connections are made progressively along it in the same order. The bus wire should be connected to the chassis at one point only, at the input end of the amplifier." (my emphasis)
The big mistake is to connect to the chassis at the power supply side instead, in which case the input will amplify all the hum/noise from the filtering current going through the bus wire. I made this mistake wiring up a tube preamp and it was very noticeable with a high level of noise and hum. So if your bus wire is connected to the chassis at the PT center taps the way I think you drew the green line, it is probably inferior. The very opposite end of the bus wire at the input jacks should connect to the chassis.

 

[ruffrecords]

I once was sent an old RCA PA amp (single ended 6L6 output plus a couple of pentodes) to modify into a compressor. What struck me most was their method of handling analogue 0V. They used a single chunky (like 18AWG) bare copper wire. It was connected at one end to the final smoothing cap of the HT supply and it extended unsupported the full length of the chassis. It was only connected to the chassis itself at one point - at the input RCA connector. The only other connection to the chassis was from the earth pin of the IEC mains connector.

Along its length the 0V connections from each stage were connected to it working backwards from the power supply to the previous stage and from there to the input stage. A couple of 300 ohm resistors were connected across the heaters of the first stage and their centre point was connected to 0V the bus wire. This was one of the quietest ac heater amps I have ever worked on. If I were to build anything point to point these days I would use that same technique.

Cheers

Ian

 

[musipol]

I've done similar ground treatments to 60-70's Fender's and it seemed to improve the hum level. What I never did was to measure the initial hum level vs the final level after the busing treatment.

 

[soapfoot]

An AB763 as implemented from the factory has typically been plenty quiet for me, especially when considering all of the other sources of noise... not the least of which are magnetic pickups (particularly single-coil pickups).

This is an interesting exercise, however. I have a Moviola URS film editing amp ("squawk box") that's been repurposed as a guitar amp, and it's quite a bit too noisy... they all are/were. I'm positive that something could be done to improve this, but the basic "low hanging fruit" I tried didn't really seem to do much (I was able to improve it by maybe a dB).

I'd like to dig in and find a way to make it quieter "one of these days"

 

[Matador]

So if your bus wire is connected to the chassis at the PT center taps the way I think you drew the green line, it is probably inferior. The very opposite end of the bus wire at the input jacks should connect to the chassis.

Yes, I didn't put this connection in a different color to make this clear to readers: the green bus is isolated from the chassis, except:

The chassis and the circuit are completely isolated from each other, with the exception of a single connection (at the vibrato input jack).

I put it there because in effect, this amp is multi-channel. However the vast majority of people I do repairs for only ever use the vibrato channel, so I thought it would make sense to go there.

What I never did was to measure the initial hum level vs the final level after the busing treatment.

What would be a fair way to test this? Perhaps with nothing plugged in (all inputs shorted), dime all of the volume controls, and take an 10-second averaged FFT direct from my load box?

 

[fazer]

Have you seen this from Martin Manning

http://ampgarage.com/forum/viewtopic.php?f=8&t=34541
Not like your ground bus but is a deluxe reverb with clean grounding.

 

[musipol]

What would be a fair way to test this? Perhaps with nothing plugged in (all inputs shorted), dime all of the volume controls, and take an 10-second averaged FFT direct from my load box?

That sounds reasonable, although I'm certainly no expert on measurements.

With respect to the DR replica build, I would prefer the gnd bus setup. The improved ground scheme shown shows a number of different chassis gnd points and it doesn't mention the input jacks being isolated. The proof is in the testing, I guess.

 

[fazer]

Martin is a very experienced tech. Very knowledgeable. Guitar amps are tricky but I see this type of grounding bus layout ( as matador is using), used in other amps and am aware of it being employed with success. Kevin O’Connor’s London amp books suggest this type of grounding. It’s easier to employ when filter caps are distributed next to the amp circuit rather than in a doghouse like the fender uses.

 

[Matador]

Ideally, each main filter capacitor should be placed directly on the eyelet board close to the preamp nodes it is bypassing. In that way, you could created (mostly) local 'stars' in a similar attempt to what was done on that DR replica build. This would keep return current to those caps limited to not flowing down long wires or long 0V busses. This would be pretty hard with an eyelet board due to the vast amount of horizontal wasted space.

However Martin's 0V tie points seem clearly bolted to the chassis, which means return current must be flowing in the chassis (although for later preamp stages the currents are likely small). I do like the idea of making A-, B-, C-, and D-, and running those to a bus however. Having a doghouse complicates things.

I don't see any advantage to Martin's (or Fender's original) scheme, other than possibly having for lower 0V impedance, due to the amount of bulk metal in the chassis. However a 12 gauge bus bar has a resistance around 1mOhm per foot, so I can't imagine it being significantly worse. Fender's scheme definitely is easier/faster to implement, as you don't need any extra wires (cheaper), and you don't have to pay someone to make bus connections (cheaper again).

 

[JohnRoberts]

What would be a fair way to test this? Perhaps with nothing plugged in (all inputs shorted), dime all of the volume controls, and take an 10-second averaged FFT direct from my load box?

The testing depends on what you are looking for. Good practice is to terminate inputs with equivalent source impedance to what it is expected to see in use. For a lead guitar maybe 50-100k ohm to mimic lead guitar pickups.

I am not a tube guy but my mixer design engineers used to share lab space with the guitar amp design engineers. I used to chew the fat with the top guitar amp designer and don't recall much hand wringing over PS noise. Some amount of hum is pretty much expected from lead guitars with massive voltage gain. That's how they tell the amp is turned on.

Physical layout was considered important for sound quality (tone) characteristics. High impedance, high gain tube circuits will often experience crosstalk between nearby circuit blocks. This can be good, or bad, or whatever. It does define the sound character of some iconic designs.

This is also the good news/bad news about PCB layouts vs point to point wiring. With PCB layouts the crosstalk will be the same every time for better and/or worse. A management dynamic from the amp design engineer's perspective, his boss is not receptive to doing one more PCB cut just to try to make it a little better, management wants to boot the amps over the wall to the factory to start building.

Today's TMI.... One of my friends was IMO Peavy's best bass amp design engineer. He actually started out working in the digital group. Back in the early days A/D conversions used DACs and successive approximation (guess, compare, repeat). On his own initiative he tweaked the layout of one of Peavey's early cheap digital reverbs with an exacto knife and wire jumpers, he managed to extract one more bit of input conversion resolution just from changing the layout. When he proudly showed his boss, what he had accomplished, his boss (director of digital engineering) got angry and told him to forget about making a running change (to avoid the corporate politics hit from a costly engineering change, and admitting the mistake). After that the design engineer transferred to the analog group and designed bass amps that didn't suck for several years. [/TMI]

JR

 

[trobbins]

Optimizing is an interesting word to use. One can optimize aspects like ease of construction, theoretical design, noise floor, hum level (measured and audible), sensitivity to different inputs causing noticeable hum, ...

Without some baseline measurements, the outcome of making certain changes may span the range from making no measurable difference to making some aspects worse (eg. due to parasitic effects that weren't appreciated). And sometimes just making sure the amp is operating properly (ie. nominal performing valves, and parts that have not drifted/degraded in performance) is the best staring point. I would expect that many commercial vintage amps could see a measurable improvement - that may be tangible in a recording situation, but may not be discernible in the intended application.

 

[Tubetec]

Sources of noise in old style tube amps often partially cancel depending on settings of the controls , I've noted that in several of the older pre CBS Fender amps . You may succeed in lowering the noise in the first stage ,but it may well not net you the gain in terms of S/N you expected due to partial cancelation later on in the circuit .

 

[musipol]

On which pre BS Fenders did you see this partial cancellation effect? Also curious which control settings?

 

[Tubetec]

Twin reverbs and other smaller amps with a pair of 6L6's like Vibrolux , even the modern blackface twin exhibits it so some extent , output stage tube mismatch can complicate matters further . For instance the hum dinger on a modern repro blackface twin reverb , optimum position depends on the settings of other controls on the amp , with the volume turned down the position for minimum hum is different to with a guitar plugged in and the volume control advanced .

 

[fazer]

A friend of mine, Lord valve, or as I Know him Fat Willie, put together a 35 watt fender style of amp. He was using the ground bus design. And said the power tubes had parasitic distortion. Finally grounded the output cathodes to the ground point lug on the transformer to chassis mount and it became stable.

But there are boutique amps that use this system. I’m not sure which ones.

 

[musipol]

Just because LV had parasitic osc. doesn’t invalidate the gnd bus approach. It may say something about the less than optimum filter cap “doghouse” layout that Fender used. I can say I have used it without parasitics quite a few times. I believe it works best with local filter cap placement forming local stars, with each connected to the bus in order.

 

[JohnRoberts]

I am probably repeating myself but I used to work under the same roof as several professional guitar amp designers. These were not casual thrown together designs but the results of experienced initial designs and test bench analysis to refine prototype results. Another observation, in my experience the better amp design engineers were also competent players. Many Peavey design engineers gigged in local bar bands, often flogging guitar amp prototypes in the real world.

I recall one bass amp designer who could not play a lick and it was painful to listen to him plunking away. In one of my early offices I was upstairs one floor above his office and even that was unpleasant. The design engineer who replaced him could really play and the product reflected that difference.

Be careful about making broad sweeping ASSumptions about the merit of one design/layout approach over another. It always comes down the execution. Details of the design matter.

JR

 

[Newmarket]

The testing depends on what you are looking for. Good practice is to terminate inputs with equivalent source impedance to what it is expected to see in use. For a lead guitar maybe 50-100k ohm to mimic lead guitar pickups.

Is that 'just' trying to match the resistance (DCR) of the pickup winding ?
50K - 100K seems bit on the high side for that. A quick look shows a Seymour Duncan High Output Humbucker at a 16K6 and a less 'hot' humbucker at a bit over 8K.

 

[JohnRoberts]

Is that 'just' trying to match the resistance (DCR) of the pickup winding ?
50K - 100K seems bit on the high side for that. A quick look shows a Seymour Duncan High Output Humbucker at a 16K6 and a less 'hot' humbucker at a bit over 8K.

Guitar pickups have inductance too... I was crudely ball parking that. A 10k resistor might understate the HF noise. To do it right maybe use an actual guitar pickup, shielded inside a metal box. For tweaking layout I don't think the 50k-100k resistive termination will steer you too far wrong.

JR