Opinions on BW-Limited Output Transformers?

[bob p]

I'm looking at transformer specs for an EL84 amp project. I'm thinking along the lines of an AC15 or a Spitfire... something with minimal preamp gain that will allow me to crank a pair of push-pull EL84 to get power tube clipping. I'd like some advice from those of you who have tried various types of iron in your builds.

There's a wide spectrum of OT iron out there, ranging from the traditional full bandwidth HiFi iron (which ends up being quite heavy), to modern bandwidth-limited iron that is specifically designed for guitar amp applications. Just as you'd expect, the BW-limited iron ends up being pretty lightweight and a lot less expensive than the full BW iron.

Some of the classic amps used better iron (larger bw) than was actually needed for the guitar amp application, primarily because surplus OT like the RadioSpares were readily available at the time and were CHEAP. These amps ended up getting heavy iron with BW capabilities that exceeded the actual needs of the application. Some might contend that the extra-heavy iron plays a large role in the amp's tone.

Traditionally, home builders in North America didn't have a lot of options for iron, and they used a lot of generic Hammond HiFi iron for their amp builds. The Hammond HiFi iron also has excessive BW for the guitar amp application -- typical Hammond ratings might be 30 Hz to 30 kHz. The result is that the iron tends to be quite heavy, is relatively over-sized for the limited BW application, handles power quite well, and tends not to get saturated.

In recent years the builders' choices of iron have expanded dramatically, as boutique iron builders have capitalized on people's desire to seek vintage tonality. In the old days I used to be able to buy boutique iron from Mercury for less than the cost of an equivalent Hammond. Back then MM iron was a good buy. But as they've expanded their marketing efforts to convince everyone that they've got something special, their prices have climbed to the point that they're now 2x-2.5x the cost of a Hammond. This is why I haven't bought MM iron in 10 years. Today it just costs too much for what it is.

As the price of iron and copper has climbed, Hammond has responded by pushing price in the opposite direction -- by producing a line of bandwidth-limited iron that's specifically marketed to the repair and DIY amp builder market. This type of iron might only be rated for 70 Hz to 15kHz, and as a result of not needing to extend the bass response, the weight of the OT is about 1/3 of what it might have been otherwise.

I've been ruminating over numbers lately, and I thought I'd ask for opinions from those of you who have tried building amps both ways. I'd be particularly interested if anyone has done direct comparisons of the heavy HiFi iron to the lightweight 1750/1760 series guitar amp iron, or if anyone has done direct comparisons of the traditional / boutique iron to the new lightweight hammond stuff. Conventional wisdom on the internet (which I believe is based upon less informed end users buying into aggressive marketing) is that you need to spend $275 on a reproduction of a Haddon, Woden or Albion transformer to get the classic Vox AC30 or AC15 EL84 tone. That's one end of the price spectrum. The new 17xx series Hammond iron is at the other extreme -- you can buy a BW-limited 1760E that's marketed as a Fender replacement for about $40. And right in the middle are the traditional Hammond HiFi OTs that cost about $100.

Looking at specs, it's hard to imagine that the Hammond HiFi iron could be at all bad compared to the (overpriced) boutique iron, or that the BW-limited Hammond "guitar amp" iron is going to be lacking either. Nonetheless, conventional internet wisdom on the guitar sites says that putting a Hammond 1760E into an AC15 build is going to make the amp sound like a Princeton Reverb or a Tweed Deluxe. (What?!?) I think that's just silly, but I'd love to hear your opinions.

Just for comparison, here are some specs:

Mercury Magnetics Boutique repro iron: no specs. $275.00.
Hammond 1608A: 8k primary, 10W, 30-30kHz +/- 1dB referenced to 1 kHz, 2.5 lb. About $100.
Hammond 1650E: 8k primary, 15W, 70-30kHz +/- 1dB referenced to 1 kHz, 3.5 lb. About $100.
Hammond 1760E: 8k primary, 15W, 70-15kHz +/- 1dB referenced to 1 kHz, 1.1 lb. About $40.
Hammond 1650F: 7k6 primary, 25W, 30-30kHz +/- 1 dB referenced to 1 kHz, 4.0 lb. About $100.

I'm really in need of some compelling arguments to convince me that the boutique iron is worth it. Or that the full BW HiFi iron is going to be much better than the 1700 series.

Your thoughts?

[pdf64]

No experience of boutique, but moving up in size from Tremolux size 022848 / Hammond 1750J to Super Reverb size 022855 / Hammond 1750M ( on a 6L6GC P-P @ 420VB+ BF Fender based arrangement) was a noticeable improvement. A bigger / deeper / fuller sound.
I don't recall the bench measurements indicating any difference though.
Regarding the weight, bare in mind that modern magnetic materials are more efficient, less iron needed.
Pete

[R.G.]

The bandwidth you're talking about is almost entirely low end bandwidth. While making a transformer bigger (i.e. with more iron) does make high frequencies harder to get because of the inherently bigger spaces in the window for flux to leak outside the coils, the biggest effect is the low end bandwidth.

Guitars do have that fundamental at 82Hz in standard tuning, but a low frequency rolloff does have some losses above the nominal turnover frequency. So even with a "70 Hz" -1db spec, there's enough loss for a truly dedicated cork sniffer to ferret it out.

I don't know any way to argue the "worth it" angle. That happens in your own ears and/or wallet.

I do have an engineer's approach to this, though. I much prefer to get as much of whatever (low end being one of them) as I might ever need, and then cut that back as needed. It's easy to choke off bass somewhere inside the amp if I like. It's hard to grow an output transformer bigger if I'm not really happy with the low end full/present/thud/body/grit/yada/yada.

If it were me - and it's not, so take that into account - I'd go for the best low end I could, and then trim it back in the preamp somewhere. Probably I'd make that switchable, too.

[bob p]

Yes, low end bandwidth. I thought that was obvious.

I'm having trouble putting faith in anecdotal experiences that maintain that there's a tangible difference that can't be measured. IMO those kind of arguments belong in the domain of the cork sniffing HiFi salesmen who earn their living selling gear by leveraging the placebo effect, and in the domain of the charlatans who want to sell you $50 hand-rolled coupling capacitors, and not in a technically oriented Theory & Design forum like this one. If someone is inclined to maintain there is a difference, I'd really prefer that the claim be accompanied by an objective quantifiable measurement.

R.G., I'd be interested in hearing you expand on the claim that low frequency rolloff has some losses above the nominal turnover frequency, so that cork sniffers can ferret out 1 dB changes in frequency response. Are you saying that there's something going on that the -1 dB figure does not measure, or that the cork sniffing types are prone to fuss over a 1 dB difference? Or was it something else?

As far as switches go, I'm all for switches, but the objective of this particular amp build is to have as few controls as possible.

[jazbo8]

Don't know if you have seen this, but some measurement and listener feedback data were presented by Tubelab awhile back. Not much surprise, the general consensus was that the larger, heavier transformers sound better, but just how much better and how they stack up on the value scale, you need to read the report and see for yourself.

Jaz

[Chuck H]

For any 2 X EL84 amp I would go with a standard Hammond 1608. I use to get a good price on these at AES but it seems the 1608A model has replaced the standard 1608 option. Still, shop around. No better sounding transformer to be had. Promise. A former member here (shea) won a couple of amp shootouts using this iron in his 18W clones. I have a proto with this iron and it sounds better than the custom Heyboer OT's I put in the production models. Don't mind the 10W rating or the convoluted secondaries. The OT can handle 20+ watts in a guitar amp easy. The secondaries can be wired for at least two selections with a standard toggle. I use to get them at AES for about $55. Now the cheapest price I can find is Mouser! (use to be the most expensive!) at $79. But a great OT just the same. Some careful shopping may reveal a better price.

[Steve Conner]

I think an important consideration when sizing an OT for a combo amp is how much bass your chosen speaker can handle. An undersized OT will choke off the fundamental and turn it into harmonic energy, abusing the speaker less and maybe even giving the illusion of more bass. (At the cost of stressing the power tubes more with a nasty reactive load.)

Filtering low end within the amp won't necessarily do this, unless you introduce distortion before the filtering.

On the other hand, if you're using a big speaker that can reproduce genuine bass, then the bigger OT will sound better.

[R.G.]

Yes, low end bandwidth. I thought that was obvious.

It was, mostly. I was saying that to make the comments about losing some treble with the increased leakage inductance from a larger core make sense as well.

R.G., I'd be interested in hearing you expand on the claim that low frequency rolloff has some losses above the nominal turnover frequency, so that cork sniffers can ferret out 1 dB changes in frequency response. Are you saying that there's something going on that the -1 dB figure does not measure, or that the cork sniffing types are prone to fuss over a 1 dB difference? Or was it something else?

The thing is, the turnover frequency is an abstraction. Frequency is not flat up to some "corner" like it's often described, then drop at a ruler-straight -Ndb/octave. What really happens is that the reactive parts (inductance and capacitance) and resistive parts of the impedances involved are always interacting, at all frequencies. The true response is asymptotic to the idealized "corner" plot.

This only becomes apparent on graphs and such at about a factor of ten around the turnover frequency, which was more or less arbitrarily defined as the half-power point. So for a half-power turnover of (for instance) 70Hz, the power lost is 3db (power) or 6db (voltage, if the impedances are equal) at that frequency. But the effects of the impedances that lead to that turnover at 70Hz are visible up to around 700Hz as a slight loss of output power.

The specification of a transformer with a -1db point at, for instance, 70Hz really means that the transformer has some lower -3db point. The math says that this is about half the -1db frequency, or in this case, a -3db point of about 35Hz gives you a -1db point of 70Hz. At some higher frequency, it's only -0.5db, and higher than that it gets to -0.2db, -0.1db, -0.05db, and so on. Only at frequencies far above the half-power point can we ignore the fractional-db drop. Where we can ignore the losses is a personal choice, because all ignorance is personal.

The hifi/tweako/cork-sniffers' union is often heard to talk about sub-db differences being audible. Maybe. If it is audible, it means (a) there is program content there to hear and (b) the speakers can reproduce it.

Steve is correct - there are psychoacoustic effects of dropping out fundamentals of a note, hearing only the harmonics. Your ear will re-insert the fundamental by beating the harmonics. And tubes/circuits may do funny things when having to drive a highly inductive load. My preference is to make the power section have as flat a response as possible over the frequency band it will get, and then to do odd filtering and other "funny things" in low power sections. That, IMHO, takes it out of the realm of adding mandrake root and stirring with a black cat bone at midnight and into the realm of deciding what to do and then doing it. But that's just me.

[bob p]

Thanks to everyone who's posted. Lots of interesting ideas.

I have always been disappointed that nobody bothers to mention L in their transformer specs.

I had trouble drawing firm conclusions from the single ended transformer evaluations at that other site. The experiment had a lot of uncontrolled variables which obfuscated the analysis for me; OT were used that had different primary impedances which caused shifts in operating points, the bias conditions were not the same for every OT (some were biased hotter than others, which also causes a shift in operating point), some OT were of dramatically different sizes. Even though the investigator allowed operating points to differ in terms of load impedance and bias settings, he concluded that those variables weren't important and that OT mass was responsible for the sonic differences. I'm not sure that's a valid conclusion.

Great info Chuck. Shea's success with the smallest Hammond challenges the conventional wisdom that bigger is always better. And your experience with the 1608 vs. Heyboer challenges the conventional wisdom that boutique, specific-purpose winds are always better than generic equivalents. It would be very interesting to see if/how things might change when iron gets used whose bandwidth is intentionally limited to the frequency bands relevant for the guitar, like the 1760E vs. the 1608.

I'd like to particularly thank those who have taken the extra step to crunch numbers to help provide a good answer. Since I'm asking these questions in the Theory & Design part of the board, I'm not going to be squeamish about math. I like math. I Engineering math. Transformer math. Derivations. Proofs. I invite anyone who's bothered to do math related to these questions to show their work, for everyone's benefit. (it wouldn't bother me to read through 6 pages of formulas if they prove a point.) Math is a precise language. Sharing math can only help us all to learn, and I'm thinking that I can't be the only person reading this thread who would prefer to solidify their understanding of the problem through math. Math is good. It frees us from having to rely upon the opinions of "experts" and "gurus".

[Steve Conner]

They probably don't bother to mention L because it's highly non-linear, especially in push-pull OTs that lack an air gap. Hard to do math when you don't have the numbers.

[bob p]

/* hijack:

Filtering low end within the amp won't necessarily do this, unless you introduce distortion before the filtering.

I guess I'm about to hijack my own transformer thread. Not to get too far off of the transformer selection subject: In this particular application, I'm planning a bare-bones preamp that will do nothing more than provide the output section with a clean signal so that the power tubes can be responsible for the clipping. Filtering in the amp wouldn't help.
*/

Regarding L: non-lineariety isn't an absolute deal breaker. It just means that they have to publish a plot instead of a number, and we have to deal with non-linear functions. More cumbersome, but do-able.

[pdf64]

Is it possible/likely for an overdriven wave to include frequencies below the input sine wave? I guess power supply ripple and the modulation effects on things would play a part.
'Yes, low end bandwidth. I thought that was obvious.'
From practical observation with a heavily overdriven amp, the SR sized OT I mentioned below had more of everything, top end included; the small tremolux sized OT seemed to have a very filtered, band limited response for both ends of the spectrum.
Postulating on that, when saturated with sufficient low end input power, the OT primary inductance would probably collapse; that would probably impact on the any other simultaneous higher frequency content.
Speaking of the math, I'm still in trauma from EM theory 30 years later!
Pete

[bob p]

A former member here (shea) won a couple of amp shootouts using this iron in his 18W clones.

Shea was a master of understatement. From the Archives: AMPAGE Archive: Re: Hammond transformer in guitar amp

Re: Hammond transformer in guitar amp

I use the Hammond 1608 (which they rate as a 10-watt output transformer) for 18-watt amps, and it sounds good.

Shea

[Gingertube]

Hammond 1608 Primary Inductance is 57H. This value is from a spreadsheet that Hammond sent me a few years back. I use 1608 in my HiFi Amp driven by 6V6G in Ultralinear.
For my 6V6 Guitar Amp I use the EL84 PP Output trannies from here:
mablexporting.com
I leave the UL taps unconnected.
My view/opinion/experience is that better output transformers help, up to a quality point, beyond which you just waste money. The cheaper 10W push pull ultralinear HiFi trannies are what I have found work well for the 14 to 18 Wattt PP 6V6 or EL84 Amps. These trannies (because of the UL taps) will have 4 section primaries and most likely 3 section secondaries. That is about the maximum level of "sophistication" which gives some benefit, offering some improvement in interwinding capacitance and leakage inductance over the "standard" guitar amp output tranny.
Cheers,
Ian
Cheers,
Ian

[R.G.]

Is it possible/likely for an overdriven wave to include frequencies below the input sine wave?

A simply-overdriven sine wave? No.

A simple sine plus harmonics? No. Any distortion produces both harmonic distortion and intermodulation distortion. Intermodulation distortion produces frequencies that are the sum and difference of the parts. With a sine and its harmonics, the sums are multiples of the original sine, the differences are the sine.

A complex waveform of non-harmonically related partials? Yes. Intermodulation gives sums and differences, and the differences get smaller than the lowest partial.

I guess power supply ripple and the modulation effects on things would play a part.

Yes. Again, IMD.

From practical observation with a heavily overdriven amp, the SR sized OT I mentioned below had more of everything, top end included; the small tremolux sized OT seemed to have a very filtered, band limited response for both ends of the spectrum.

The problem with that observation is that it's not controlled for the effects of leakage inductance. Leakage inductance is one of the primary limiters of top end, and within certain bounds, it's independent of primary inductance, which is the limiter at the low end. You have to test both, and also look at the ratio of primary and leakage inductance. The ratio Lp/Ll was used in the Golden Age as one of the primary measurements of the goodness of an OT.

Postulating on that, when saturated with sufficient low end input power, the OT primary inductance would probably collapse; that would probably impact on the any other simultaneous higher frequency content.

Depends, depending on what you mean by "collapse" and by the ratio of the low to high frequency content.

For normal transformer irons, saturation is not a knife-edged event. At some point, the incremental primary inductance to the next unit of drive voltage X time starts to decrease. And it's an instantaneous event, so the transformer is only "saturated" to extent it's saturated, for the time it's over the sloppy, indistinct knee of the start of saturation. In that range, Lp to any additional signal volt-time added decreases. How much it decreases depends on the size and frequency of the added signal. For signals of 100x the base frequency, the Lp may have decreased a lot, but not enough to make the 100x of the frequency ratio have problems from saturation.

Saturation is a low-frequency effect. Iron saturates based on the integrated voltage-time product. So if a transformer saturates at the volt-time equivalent of 50W at 100Hz, it won't saturate til 100W at 200Hz, 200W at 400Hz, 400W at 800Hz, etc. Other losses come in at higher frequencies, but saturation isn't one of them. This is why 100W power transformers can be smaller than your thumb at 1MHz.

What it might do is quiet any high frequency content in the next couple of octaves above the saturating signal, and then only at peaks of the saturating signal. This would be more like a tremolo at the low frequency, which would sound ring-mod-y at these frequencies. Signals at more than a couple of octaves up would be much less affected.

Magnetic amplifiers have always been a fascination of mine. These work best if made from "square loop" magnetic material where the transition from unsaturated to fully saturated is as abrupt and sudden as the iron mongers can make it. The gain of the mag amp is dependent on how suddenly the iron can go from unsaturated to saturated.