ST-70 hotrod header
ST-70 dragster 850

The amazing Supermods™ technique can double the output power of almost any push-pull, fixed-bias, hi-fi tube amplifier! (Assuming it has space for more tubes.) Uses the original transformers and works at very low and high audio frequencies, too!

Reported by Stephen H. Lafferty

Note: This project pushes the power and output transformers of the ST-70 beyond their specs. While we are not aware of any ill effects, we cannot be responsible if any occur.

When I first saw the posting by Andres Fontanills in the diyAudio forum, I was incredulous. With his Supermods™, he was claiming to be getting 71-watts per channel out of the same output transformers from which I had seen only 35W. Perhaps more amazing was the fact that he measured 53W out at 20Hz, whereas I had seen only 19W (with our ST-70A mods), seemingly limited by saturation effects. How could this be? Well, it can be, as we will show. We contacted Andres and he was very gracious in working with us on a seemingly endless series of questions and tests to verify his measurements. He seems like a very nice and earnest fellow, pursuing his vintage audio hobby on limited resources (as most of us are). Below are his measured results (8-ohm load), taken just under visual clipping (more about that later):




















62W, both chan driven


*Stock power was tested at 1% distortion, whereas the Supermods case was tested at visual just-below-clipping. Generally, Andres’ max level pictures look better than 1% clipping on a scope. 

How it Was Done

In a nutshell, four tricks:

  • Operating the output transformer at half-impedance.
  • Adding four 7591 output tubes, in push-pull-parallel.
  • Using a solid-state rectifier and adding a small transformer to power the new 7591 heaters.
  • Using extraordinary negative feedback to reduce distortion from transformer saturation.

Eico ST-70 with Supermods iconAndres used some of the modifications given in our original article about improving the Eico ST-70, reducing distortion. Hereafter, we will refer to that article and those mods as the ST-70A. Also, he made some changes to the feedback and frequency compensation of the amp. The schematic of his power amplifier is in the thumbnail at left. This is based on the Sams schematic, so component names do not match Eico’s schematic. Note: The original Sams schematic erred by interchanging the BLK and YEL color labels on the output transformer leads, corrected here.

Half Impedance and Push-pull Parallel7591EH low

Let’s review the changes. Operating the output transformer at half-impedance means using the 16-ohm tap for an 8-ohm load, the 8-ohm tap for a 4-ohm load and the 4-ohm tap for a 2-ohm load. The load seen between the plate connections on the transformer will then be 3250-ohms, which is half of normal. To drive this load, a new 7591 tube is added in parallel with each existing one. Andres used Electro Harmonix 7591’s (right). You can read about our tests on various 7591 tubes at this link.

Provisions must be made to handle the fact that the tubes require different amounts of grid bias. Each channel of the original ST-70 circuit had one pot for output tube balance, and one pot for overall output tube bias. Andres leaves this in place for the existing two output tubes. For the two new output tubes in each channel, he has added two pots, one to adjust each tube. This lets you adjust the balance of the two new tubes and also balance them against the original pair. 

Solid State Rectifiers and Auxiliary Heater Transformer

ss rectReplacing the original GZ34 rectifier tube with solid-state rectifiers (at right) does three things:

  • Supports the higher current demand.
  • Increases the DC power supply voltage.
  • Relieves the power transformer of supplying the rectifier heater.

Of course, higher current is demanded by the doubled output tubes and the halved load impedance. The original GZ34 cannot support that. Aux xfmr lowThe increased DC supply voltage, from the solid state rectifier, contributes to extra power output. Relieving the rectifier heater load helps compensate for the higher DC load on the power transformer. To keep the ripple constant, the first filter cap gets doubled, too.

Rather than power the four new 7591 heaters from the existing power transformer, Andres added a small filament transformer (at left) to take care of that. This is important to preserve the original power transformer capacity for the DC supply. The new transformer must be able to supply 6.3Vac at 3.2A.

Increasing Negative Feedback, Reducing Gain and Distortion

The changes discussed so far enable the amplifier to push more power through the existing output transformers. However, particularly at low frequency, there is no doubt that we will be operating it further into the nonlinear saturation region than before. This would generate more distortion if it were not for the increased negative feedback provided in this project.

Negative feedback reduces the overall gain of an amplifier (hence the term, “negative”). The more negative feedback you have, the less the overall gain will be. One benefit of higher negative feedback, is lower distortion. The added compensation components around V4 came from the ST-70A article. Speaker Terminal wiringThat allowed us to increase the negative feedback by 10dB, re­duc­ing the gain and distortion of the power amp by the same amount. Andres is also using the new feed­back components of the ST-70A, with a couple of exceptions: (1) Instead of a 4.3K feed­back resistor, he has a 10K pot at R37 (at right). (2) Instead of grounding the black trans­form­er wire and taking feedback from the orange wire, he grounds the yellow wire and takes feedback from the green wire (as Eico did).

This has the effect of reducing negative feedback by 3dB, relative to the ST-70A. However, this could be more than made up, using the variable feedback resistor. The ST-70A article did not change the overall gain of the unit, as reductions in the power amp were compensated by increases in the line stages.

However, we have since noticed that the overall stock ST-70 gain seems somewhat excessive and could stand to be reduced by reducing the power amp gain. This may be done using the variable feedback resistor.

On the other hand, the power amp could become unstable at very low gain, having too much negative feedback. It seems to us that this pot should be adjusted with some consideration given to stability. That could be assessed by testing with low level square waves, both at high frequency (say 10kHz) and ultra low frequency (say 0.5Hz), using a DC-coupled scope.

In addition to the selected ST-70A mods, Andres has added RC networks across the primary of the output transformer, to improve stability.

Option for the Neg FB Adjustment and Other Items

Andres made the negative feedback of the power amp adjustable, because the greater power output capability might require greater gain. Personally, I would go with a fixed feedback resistor. Andres is open to that idea. Along with that, I will suggest a couple other items and the reasons for them (click illustration, below right). It would be best to either take all three parts of this option or none. In the section on making the mods, you will be able to choose which way to go, as both will be presented.Alternative NFB and gnd icon

  • The first suggestion is to used a fixed, 4.3K resistor in place of the Neg FB pot (R37). This avoids the need for adjustment and any need to check stability. It is the same value as used in the ST-70A article. Used with the next suggestion, it results in an overall 3dB gain increase over the ST-70A (and ST-70). That is commensurate with the doubled output power provided by the Supermods™ technique.
  • Instead of taking the negative feedback from the 8-ohm (formerly 16-ohm) tap of the output transformer, I would use the 4-ohm (formerly 8-ohm) tap. This was the same tap used for the ST-70A, so tap-related stability issues are well-addressed by the compensation mods. Used with the 4.3K resistor for R37, it provides the same amount negative feedback which worked well in the ST-70A article and thus should avoid any stability issues.
  • Instead of grounding the 2-ohm (formerly 4-ohm) tap, as Eico originally did,  I would ground the transformer Common wire. This change was also used in the ST-70A, so keeping it the same is part of insuring stability and realizing the gain and feedback levels, mentioned above. A key advantage of grounding the Common, is that the speaker grounds may be connected and their ground may be connected with the input ground. This is very useful for testing, speaker switching and headphone connections. However, it does eliminate the center speaker connection feature (rarely used).

Using these three suggestions avoids the NFB Adjustment, insures stability, provides 3dB of gain to support the extra power, keeps the high (30dB) negative feedback provided in the ST-70A mods and it allows the speaker and input grounds to be connected. However, it does depart somewhat from Andres’ project, so there is always some possibility that the results won’t be what you expect.

Making the Modifications


Here is a list of the steps which are needed to complete this project:

  1. Implement the mods from the ST-70A article. Andres didn’t use the AC balancing part and the phono preamp part isn’t relevant to this project.
    • If you choose the option discussed above (see Alternative Negative Feedback sketch), install the 4.3K resistor instead of the Neg FB pot, take the negative feedback from the 4-ohm (formerly 8-ohm) output and move the output ground to the output Common of the transformer terminal board.
    • If you do not choose the option, instead of the 4.3K feedback resistors used in the ST-70A article, install the 10K pots, which Andres used. Output ground remains at the 2-ohm (formerly 4-ohm) output and negative feedback comes from the 8-ohm (formerly 16-ohm) output.
  2. Add the four 7591 tubes to the chassis. This may require that the existing tube, V2, be moved. The octal sockets for the 7591 required 1-inch holes and the 9-pin socket for V2 needed a 7/8” hole. Using chassis punches is the usual way to make the holes but Andres used spade bits, intended for wood! He recommends a slow drilling speed and cutting oil.
  3. Add the four bias pots and associated resistors and caps for the new tubes. Optionally, add the four bias test point banana jacks.
  4. Add the four RC networks which are shown across the primaries of the output transformers.
  5. Remove the GZ34 rectifier tube and install the solid state rectifiers.
  6. Add the 40uF filter capacitor across C1A.
  7. Add the 10K 1/2W resistor across R98.
  8. Relabel (or remember) that the 16-ohm output becomes 8-ohms, 8-ohm becomes 4-ohms and the 4-ohm becomes 2-ohms.
  9. Adjust the DC bias controls using the procedure, below.
  10. If you did not take the option (alternative negative feedback) discussed above, adjust the Negative Feedback control using the procedure, below.

A Tour of the Construction

At right, we have the ST-70, before the mods. The detail below shows the tube (V2) which was moved to make space for the 7591s.

ST-70 before mods low
ST-70 before mods low detail

At right, you see the final position of V2 (empty) and the 7591 which took its place, by the yellow jack (bias TP).

Final position of V2
under the chassis

Along the top row at left, we have the four new 7591 sockets, seen from under the chassis, before wiring. Banana jacks for bias test points are along the very top. The black rectangular boxes with long wire leads are the pots for bias adjustment.

Spade bit Bias and Test points low
Final Amplifier

Above left, Andres drilled the holes in the steel chassis with a spade bit, intended for wood!

Above, we see the new bias adjust pots, test points and 7591 sockets.

At left, is the final layout of the amplifier. The new 7591’s are along the left side, with heaters powered by the Aux Xfmr. The GZ34 is replaced by silicon rectifiers. V2 is moved.

Adjustment Procedures

Output Tube DC Bias

  • If this is the first time the new circuit will be adjusted, preset the bias controls for maximum negative voltage on the grids. That will insure that initial currents do not exceed safe values. After only a few minutes of warmup do an initial setting, then allow the regular warmup.
  • The amplifier should be allowed to warmup at least 30-minutes before adjusting DC bias.
  • Adjustment of the original Output Bias and Output Balance controls is unchanged with the mods but they should be rechecked. That adjustment affects the new output tubes, so it should be done first. The voltage on Pin-5 (the test point) of V5, V6 (V10/11 for channel-2) with respect to ground should be 0.38VDC. Use the Output Balance control to equalize the two tubes and the Output Bias control to set the target value. An easy way to equalize two voltages is to put the meter between them and adjust for zero. The controls may interact, so go back and forth until proper bias is achieved.
  • Comment: Unlike the original bias controls, each of the new 7591s has its own bias pot. These bias pots are mostly independent of each other but are affected by the original Output Bias controls. That is why V5/6/10/11 should be adjusted before V5B, V6B, V10B and V11B.
  • Use the V5B Bias, V6B Bias, V10B Bias and V11B Bias pots to adjust Pin-5 of their respective tubes for 0.38V. The controls may interact, so iterate until proper bias is achieved. Recheck V5/6/10/11 after this is done.

Negative Feedback (Neg FB) Adjustment

Andres recommends the following procedure for each of these pots:

  • Preset the Neg FB pot to the center of its range.
  • Connect an audio generator, with the output set to zero, to a line input.
  • Connect an 8-ohm load on the output of the ST-70 (was the 16-ohm terminal).
  • Select the input, set Level to max, Bass to center and Treble to center.
  • Set the generator to 1kHz, 360mV RMS.
  • Adjust the Neg FB pot so the output is just under clipping.

Continued on the next page:
Can the Power Transforme
r Take It?


Reader Comments

Posted by Steve L. November 13, 2019 - 10:51 am
Hi Gordon, I really appreciate your post about all the work you did on those three amps. It's very encouraging that you saw 55W at 20Hz, assuming that the total distortion was low. I'm curious whether this confirms Andreas' performance result for the ST-70 power and output transformers but I couldn't quite tell whether that was the configuration which produced your figures, since you also mentioned the monoblocks with Dynaco power transformers. By the way, perchance, are you the same "Gordon" I knew here in the Atlanta area many years ago? As I recall, that fellow had a particular interest in sweep-tube output stages with the screen grids driven instead of the control grids, at the time. If you are local, please email me at the address given in the section linked from the About button on the home page. Thanks and cheers.

Posted by Gordon W. November 13, 2019 - 10:51 am
Steve- I can specifically comment on the use of KT88s and 6550s with the ST70 output transformers in "half-impedance" mode. I have built three stand-alone power amps, two with the original ST70 power and output transformers in a stereo amp configuration (also using an auxiliary heater transformer like Andres did), and one pair of monoblock amps with an ST70 output transformer and a Dyna Mk III power transformer in each amp (using a 5U4 rectifier instead of the GZ34, to reduce B+ to a proper level for this configuration). I redid the feedback (due to the lower gain and higher drive requirements of the 6550/KT88 compared to the 7591), to retain about the same feedback (I measured it at about 28dB on mine- which gave an input sensitivity for a power amp, of 1.3v- by comparison, a Dyna ST70 is 1.4v, so this is a good range for any preamp), and I beefed up the front end and inverter supply B+ to better handle the voltage swing needed by the KT88 input grids (about +- 35v, whereas the 7591 needed about +-21v or so). I wound up using 400v to the inverters- which resulted in about 135v on the cathode, and about 135v across the plate resistor. This gave about 100v inverter swing- more than enough to provide 35v to the grids. The amp was able to produce over 50w/ch broadband at any frequency- probably more, but I stopped measuring when it just barely started to, visibly, on a scope, flatten the tops and bottoms of the waveforms at about 55 watts at 20Hz. I used about 460v B+, and about 325V on the screens, so the screens were relatively "loafing". These amps have a very open, articulate, dynamic sound, with quite powerful and controlled bass. I actually directly compared one of these, with a VAC Signature 200iQ power amp ($14500.00), using a VAC Renaissance V preamp to drive both amps- and the Eico-based amp held its own better than I expected. Yes, the VAC had a bit better imaging and soundstaging, and was ever so slightly more dynamic even at low volume- but it took that back-to-back comparison to notice any lack of anything in the "ST70 KT88 Hotrod" amp. I expect I will probably be building more of these,in the future...

Posted by Steve L. August 02, 2019 - 09:15 pm
Hi Vern, Thank you for your comment but a single pair of 7027A tubes would not support the half-impedance operation of the output transformer which is the key to the Supermods(tm) technique. The 3300ohm primary impedance is what allows the amp to deliver 71W from a single channel into the load. That means each side of the primary sees a load of just 825ohms. If we estimate the plate voltage at 400V under load and allow 50V drop across the tube, the peak current required is about 400mA. Estimated screen voltage under load is about 320V, so the 7027A could only pull about 250mA. And this assumes that the driver could deliver the higher voltage swing required by the 7027A. Recall that the 7591A uses a lower grid bias voltage and delivers max output with less drive than the 7027A.

But like you, I've also wondered about using a pair of beefier tubes instead of the push-pull-parallel (PPP) 7591As. A pair of 6550s might cover the plate current but would probably require more drive than 7591As. Would have to look into whether the driver circuitry would be up to it. There is also the usual gotcha of screen dissipation. A parallel pair of 7591As allow up to 12W peak, whereas the 6550A offers 10W. In any case, I appreciate your highlighting this interesting angle on hotrodding the Eico ST-70!

Posted by Vern August 02, 2019 - 03:35 pm
Its Much Easier to Use a Single Pair 7027A PP Tubes that can handle the Load to 75 watts and a reduced heater current. This removes the need for extra sockets and related parts duplicated.
Plus distortion is less than parallel tubes.

Posted by Steve L. November 28, 2012 - 07:49 pm
Hi Francisco, Thank you so much, for the kind words. I hope that your electronic projects are enjoyable and go well!

Posted by Francisco Gutierrez November 28, 2012 - 06:40 pm
There's only one thing that I missed to tell : I'd love to be reading this on Popular Electronics or Radio Electronics. Thanks Guys!

1 - 2 - 3 - next> - last>>

Add your comments here...