"Light Loading" Amps - Music Refence and others...

On another thread, a discussion started regarding light loading amplifiers. Since it was a discussion on Vandersteen speakers, I thought it merited an new thread, especially since there is some difference of opinion.

The principle of light loading was prposed by Roger Modjeski of Music Reference.

He said (with reference to his RM10, but applicable to all his amps):

"The amplifier is flat within 0.1dB and has low distortion of 0.3% when played below clipping on average level material. At the recommended bias current of 30mA/pair, the idling dissipation is nine watts or 75% of the tubes' rating. I estimate tube life to be 5,000 to 10,000 hours. Although higher idling currents will reduce distortion, it can also be reduced by light loading. Basically, light loading reduces the output current demand on the output tubes, allowing them to be more linear. It also reduces noise, raises damping factor, reduces distortion by 78% and allows for 80% more peak current when needed. The only loss is about 20% of the power rating or 1dB." Light loading means connecting the speaker on the tap that's one half its nominal impedance rating (i.e. the 4-ohm tap for 8-ohm speakers). For 4-ohm speakers, the he recommends running two RM-10s bridged to 70-watt monoblocks.

In the aforementioned thread, Ralph Karsten of Atma-sphere said:

"If you use the 4 ohm tap on an amplifier with a speaker of higher impedance, the output transformer will be inadequately loaded, and so it will express less of its winding ratio and more of its inter-winding capacitance. This can result in the amplifier no longer having flat frequency response. In addition, the transformer can 'ring' if inadequately loaded, which is another way of saying that it will distort.

The Merlin is an 8 ohm load, with a dip to 6 ohms or so. Its a benign load and an amplifier with an output transformer, if the transformer is designed properly, will likely work best on the 8 ohm tap. This will minimize the artifact of the transformer."

Two views. And different views from listeners, somew thinking light loading works magic, other saying differently.

What do you think of the priciple. The two technical arguments? Your experiences with light loading?
Post removed 
My experience mirrors that of Ralph Karsten. Using the lower transformer tap is typically higher distortion, more compression and frequently less bass.

Viridian makes excellent points about speaker impedance swing, it's a big factor.

I've not heard a RM10, so perhaps there is something going on with it's design that changes the rules. Or perhaps this somehow relates to the amp designers decision as to what's the perfect number of transformer turns for his creation.
Yes, Roger does mention: "It also reduces noise, raises damping factor, reduces distortion by 78% and allows for 80% more peak current when needed. The only loss is about 20% of the power rating or 1dB."

In my case my speakers Are nomnal 8 ohms and never below 6.5 ohms.

I'm interested in the variance in views between Roger and Ralph on the issue. It might be that Roger's recommendation only applies to his amps because of the way he designs the RM10s, and Ralph's take is true for most other amps.
Post removed 
I suspect you are right, because I have to believe Ralph very well knows what his talking about. It was interesting to read years back on Audio Asylum how many folks simply did not believe Roger's specs for the amp - of course he did do what others thought impossible.
Just to add to the conversation, Ralph added this comment in the Vandersteen thread I alluded to:

03-03-11: Atmasphere
Pubul57, the act of increasing the bias on the tubes has the effect of reducing their output impedance. This will change the relationship they have to the output transformer.

In addition, you have to also consider how the amplifier was designed. For example, many power tubes want to see about 3000 ohms plate to plate. But what if you had the transformer designed to be 3500 ohms plate to plate? You might loose a little power, but now you can experiment with different taps to affect the sound in different ways. This is because the transformer really does what it is called- it transforms impedance. The load that is on the output taps will affect the load that is on the tubes. Especially in the last 30 years, its been a good idea to build in a little reserve to deal with the many 4 ohm speakers out there.

So if in the above case you are loading the 4 ohm tap with an 8 ohm speaker, the result is that the tubes might see a load impedance that is much higher, perhaps 6000 ohms. Now in some cases the amplifier will not be able to make as much power, as the voltage that needs to be made across such a load to get the power is no longer available. But it may not matter if the power thus obtained is sufficient.
Roger Modjeski: The only loss (from light loading) is about 20% of the power rating or 1dB.
Hi Paul,

As we had discussed some time ago, this statement by Roger strikes me as incorrect for most amplifier designs. I have no way of knowing how correct it may be for his own designs.

In general, connecting an 8 ohm load to the 4 ohm taps of an amplifier may reduce maximum power capability by up to 50%, compared to what can be delivered into the same 8 ohm load connected to the 8 ohm taps, and compared to what can be delivered into a 4 ohm load connected to the 4 ohm taps.

A loss of about 50% will occur in cases where maximum power delivery into a 4 ohm load connected to the 4 ohm tap is limited by voltage swing capability, and where the design is such that the voltage swing capability on that tap does not increase significantly as a result of the 8 ohm light loading.

Concerning your basic question, all of Ralph's comments make complete sense to me, and I would expect them to be true. However I would imagine that in some cases, presumably including Roger's designs, the transformer may be over-designed (or designed with the possibility of light-loading specifically in mind) such that the concerns Ralph raised would not have too much quantitative significance, and/or may be outweighed by the benefits Roger enumerated.

I third Viridian's excellent comments about speaker impedance swings, and about the need to try both connections and see what sounds best.
Your experiences with light loading?
FWIW, my speakers are very close to 6 ohms throughout almost all of the spectrum, rising to around 10 ohms at deep bass frequencies. I tried both taps on my amp, and settled on the 4 ohm tap, although the difference was surprisingly small.

Best regards,
-- Al
Hi Al. To tell you the truth, I find they sound very, very similar both ways, but was predisposed to prefer the "light load" for the "benefits" in measured performance(but for power)and longer tube life that Roger mentions in his comments. I had no doubt Ralph was right as a general principal, and I do wonder if Roger just took an "roger" approach, as is his way, that made his recommendation valid - for his amp.
There is clearly something going on with that amp that allows it to output 35 watts from a pair of EL84s when no other amp is even close to the figure. In fact, there are no other EL84 amps that will make more than half that power from a single pair of EL84s. Modjeski is a brilliant designer and perhaps his point applies to his unique topology used in this amp.

Yes, a genius designed it. Three things the man is known for are: quiet amps (quietest tube amps I have heard), deriving more power from tubes than others have, and electrostatic speaker design.

In all seriousness though Roger specifically states he wrote his own application for the EL-84 tube. The original concept for designing the amp was to power his Quad ESL 57 speakers (not many know this but Roger cut his teeth working for Harold Beveridge designing the ESL amp for those speakers). Clearly the typical EL-84 application wasn't going to cut it.

For those inclined, the 6 Moons review of the RM-10 MkII is a good read with interesting comments from Roger, including the one below, as well as theories on light loading and 2nd harmonic distortion. I'd also recommend reading the RM-10 MkII manual for additional insights from Roger.


"Over the last 10 years, I have written several new applications to get higher power out of some common tubes that I thought could do more. If I were employed by Sylvania in the 1950s-70s, I would have done the same thing. There are an infinite number of applications that can be written for any given tube. They involve variations in B+ voltage, current, grid drive, screen voltage and load impedance. The data books list only a few to give amp designers some ideas. Sadly, that's as far as many designers go. David Manley was a stickler for applications believing that the only ones allowed were the ones in the data books and they must be adhered to the letter. Sadly, he took the ambitious 560-volt ultra-linear application from MOV as reliable and made some amplifiers that damaged a lot of tubes. Not even MOV could make tubes that consistently held up in that application. To this day, there are no tubes that hold up well at those voltages on the screen.

I research my applications very carefully with dissipation studies on all elements. And, although I have a replacement tube business, I do not want to waste valuable tubes (or your money) on poor applications. One application that caused a lot of comment in the Asylum was my getting 35 watts out of a pair of 6BQ5/EL84s [the amplifier under review - Ed]. They said it couldn't be done yet it is done and has proved reliable over the past 10 years in the field. The RM 245 uses a new application that yields 4 watts at a dissipation of less than 8 watts. There is no extra stress on the tube. Typical applications are 10 watts dissipation (the max for that tube). If you look at the following link, you can see the power vs distortion for 3 tubes at 25 mA. and the EML at elevated current where we take advantage of some of its extra dissipation. I have not yet written the maximum application for that specific tube but I expect to get 8-10 watts at low THD.

The above referenced graph of power output vs THD is at the load. You can add about 20% at the plate. The commonly used application from the RCA manual (RC-11 pg. 86, 1933 and repeated thereafter without modification) gives 2 watts at the plate with unspecified THD (shall we assume 5-10%). Two watts was more than enough power for the times and it seems little has been done by modern 45 amplifier designers to consider new applications.

Applications once written are forever left to rest and copied verbatim by every other maker. Remember, an application is often written to entice the designer (of a table radio at that time) into a circuit that fits his technical and economic needs. The low B+ (275V at 36 mA = 9.9 watts dissipation) of this application is attractive. Note they have 2 even lower power ones for those more economical who desired less power back then. This tube was designed to be a step up from the type 71 (max pwr. 0.79 watts) and a practical alternative to the dismally inefficient and hard-to-apply type 50. Note that the 50 was rapidly pushed aside for the much more popular 45. I have a Brunswick Panatrope that uses a 50. It's a monster that sold for the price of 1.5 Model T Fords in 1928. The 50 also had to be transformer-coupled when resistance coupling was coming on fast. My application is 310V at 25 mA =7.75 watts along with circuit and load differences.

I "light load" my RM-10 and by default have no choice but to use the 3 ohm tap on my Music Reference EM-7 monoblocks (my speakers are 8 ohm minimum with 12 ohm peaks). I'm not hearing anything negative like lack of bass, quite the contrary with the RM-10, I thought the bass improved. Having met the man a number of times when I lived in Santa Barbara I can vouch for his engineering knowledge. What he does to make his equipment sound so good is beyond me though. I just know it's relegated some really good equipment to the closet.
I suspect that Roger has a really robust driver circuit in the RM-10. IOW, its probably a class AB2 circuit.

Depending on the way the transformer is spec'ed out, I could see a loss of 50% or a loss of 20% of power if you 'light load' the amp. We build a guitar amp that uses a pair of 2A3s in push pull for about 16 watts (and is otherwise a clean version of the Marshall Plexi for you guitar nuts); loading on the amp has to be correct for the amp to make full power.

I don't have any problem with anything I have read here or on the other thread so far (I just geek out on it is all...). There are a lot of variables in an amplifier design, the transformer being a rather obvious one. But I have found that just because one designer says you can't do it is no panacea for it not being possible :)
I suspect that Roger has a really robust driver circuit in the RM-10. IOW, its probably a class AB2 circuit.

You are the second credible source that I have heard state that. The first source found reference to this possibility in the RCA book.
OK, for us non-geeks, what's an AB2 class amp and more importantly how does in vary in factors significant to human music perception from Class A or AB amps.
In this thread, Roger says the RM-10 is a class AB1:

Michael (Swampwalker), the "1" and "2" suffixes are applicable to tube amps only. "1" indicates that the grids of the output tubes are always negative. "2" indicates that they can go slightly positive, causing grid current to flow. That in turn results in some increase in output power capability, at the expense of an increase in distortion.

Best regards,
-- Al
Post removed 
Maybe we can save Ralph the trouble since you actually addressed the SA-4 here:


To quote:

03-05-06: Viridian
These were designed by Roger Modjeski for Counterpoint. With a sympathetic load, they are among the finest amps ever made.
Post removed 
There seem to be an awful lot of folks running RM9s for many years for them to be too unreliable. Now Roger did come to think of tubes on boards as being problematic over time so he went to point-to-point wiring on his RM9SE and the RM10s and 200s. Another thread questioned the reliability of Quicksilver amps to which I could only say that all manufactured goods can have a problem form time to time, but I certainly think that at least in my experience, Music Reference, Quicksilver, and Atma-sphere all are very reliable in the aggregate experience of owners.

I never heard the RM9, I owned an RM9 Special Edition and now the RM10 MKII - they are certainly quiet, but I have 89db speakers not 104db, which would proabably be the ultimate test.
I can concur with Pabul57 regarding the noise floor of the RM-10. Virtually non existant driving a pair of 98db sensitive Tannoy Westminsters. Certainly the quietest tube design I have ever heard.
Eccletique, have you tried light loading? I assume, don't ask why, the Tannoys are nominal 8ohms?
I would describe the RM 10 - to steal a phrase... "starless & bible black". Dead quiet background, resolved, tone rich & fleeting fast. It commands your attention & just draws you in. The best audio purchase I've made to date. When I received my new RM 10 from Roger the 4 ohm taps were open where he had personally burned in my unit & listened to my unit for a week & I followed suit. Definitely the way to go.
Pabul57. Yes, the Tannoy Westminster is nominally an 8 ohm load, and never drops below 6.5 ohms.
Although I do not employ the RM-10 with the Tannoy, it is one amazing little amplifier regardless of tap selection.

Pehare. Congrats on your new toy.
The Merlins VSMs are similar - 8ohm, 6.5 ohm minimum:) I
agree with you abut the RM10, it is one hell of good amp,
good enough that I sold the RM9 Special Edition ($10,000 -
162 watts!) While the RM 9 SE can drive many (any?)
speakers the RM10 might not be up to, but with the Merlins,
the little amp does the job without apologies. Combined with
an Lightspeed Attenuator, RM10 MKII, and 89db speakers with
smooth benign impedance curves, you have one heck of great
pre/amp combo capable of SOTA sound for $2,500 MSRP.
yes Paul, lite loaded for sure....they are a wonderful combo w/the RM10. I've tried & owned some other tube amps that sounded wretched in comparison w/these speakers. They don't suffer fools lightly.

Coincident Speaker Tech Partial Eclipse II
Frequency Response: 30hz - 30khz
Impedance:8 ohms ( min. 7.9 - max. 10 ohms )
Sensitivity: 92 db - 1w @ 1m
w/extender feet & oversize spikes
It seems as if most owners have found light loading this particular amp works as Roger suggests it should - better. My take away is also that light loading is unlikely to be the right approach with other amps, for the reasons given by Viridian, Albert, and Ralph. Thank you for your thoughts.
Yup, the little RM-10 can go "Toe to Toe" with many of the highly regarded heavyweights out there.

Pehare. It will get even better after a couple hundred hours of seat time.
With respect to the light loading thing..... each to his own I suppose.
After reading some of the recent comments on the RM-10, went back to the 4 ohm tap for a 6 hour session last evening.
Sorry gang but the 8 ohm tap is the clear winner over here.
Seems to me Scott made an EL-84-based amp that made 25 watts. It too was class AB1. To get 35 watts though would require one to jump through a lot of hoops design-wise (most EL-84 designs don't make more than 17-20 watts, but a good portion of those are class A). Something will have to give somewhere. I suppose if you ran enough B+ you could get that kind of power without running grid current, but OTOH you might have problems with tubes holding up, and it would not surprise me that the base of the tube was putting a lot of heat into the tube socket.

This latter issue may explain why the move away from circuit boards. But I would expect that the tubes may well have to be pre-selected, or certain versions of the tube may have to be avoided.
Hi Ralph,

Based on a quick look at the schematics at HHScott.com, it appears that the only Scott amps that used 6BQ5's were the 222A and 222B integrateds. The 222C and 222D used the generally similar 7189, as did the 299A and 299B integrateds.

This link for the 222B indicates a continous power rating of 13W/channel.

I have some older literature which indicates that the 222C, using the 7189, was rated at 24W/channel, "per the IHFM Standard." Not sure if it would rate that high under today's standards. Also, according to my tube manual the 7189 appears to be rated for significantly higher plate and screen voltages than the 6BQ5.

Best regards,
-- Al
Not only 35 watts, but Roger claims 10,000 tube life on the EL84s. He also claims light loading will extend tube life.
Haven't seen you around since the Supratek days. Hope you have been doing well. Bob
Hey Bob. Life is grand over here. Hoping things are good for you and your loved ones.