Cortina 3070 low3

A Little Love for the Cortina 3070






Performance Data and Comparing Modified with Unmodified

In reviewing the measurements, we will be comparing the modified power amp on the left channel with the unmodifed one on the right channel. The only modifications to the left power amp were the replacement of output and driver transistors, along with the addition of stability caps to support those. However, the new transistors brought a new level of performance to that channel.

Ideally in a stereo amp, you would want to keep the channels as symmetrical as possible. So it would normally make sense to go ahead and replace the transistors on the right power amp, as well. However, this amp is experimental and I wanted to leave the original side for comparison.

The final test data on the Cortina 3070 is in the table below. Shading separates grouped items.

Discussion of the Power Results

The modified left channel is generally stronger but interestingly, not much so, for the odd case of an 8-ohm load, one channel driven, at lower frequencies.


(modified PA)

(unmodified PA)

Power/chan, both driven, 1kHz, 8-ohms, 1% dist






4-ohms, clipping



Power/chan, one driven, 1kHz, 8-ohms, 1% dist









4-ohms, 1kHz









Distortion, 6Vrms out, 1kHz









Frequency response flatness, Line input, 20-20kHz

+.14, –.15dB

+.25, –.20dB

Signal-to-Noise Ratio (SNR), Line input, shorted4



RIAA response flatness, Phono input, 20-20kHz5

+.41, –.33dB

+.27, –.14dB

Signal-to-Noise Ratio (SNR), Phono input, shorted6



Phono preamp gain, at tape output



Phono input impedance, 100Hz-20kHz

50K ±1K



  1. 18.2W at 1.4% distortion.
  2. 14.9W at 1.8% distortion.
  3. 19.4W at 2.0% distortion.

  4. Volume adjusted so 0.5V in gives 12Vrms out.
  5. Measured at tape output.
  6. Relative to 10mVrms input, 400Hz to 30kHz.

Cortina 3070 power amp simplifiOne thing is clear, the new driver and output transistors really make a difference. The reason is that the 4-8-ohm load on the output appears at the input of the driver (ZL at right) as the load multiplied by both betas (the driver and the output transistor). The gain of Q301 is proportional to that load, hence the negative feedback is proportional, too. So, for example, if you double both betas, feedback goes up 4X and distortion drops 4X. By the way, R306 doesn’t get into it much because, it is bootstrapped-out, as covered in the description of how the power amp works.

From this, you can see that, for 4-ohm loads, the old, lower-beta transistors are at a real disadvantage. There’s more: The beta of the old transistors drops rapidly at high currents, so just when the betas are needed the most, it fizzles. The 4mA or so of current running in Q301 might not be enough to drive the output to full swing, or, if it’s close to max, distortion increases. Limitations on the power then, can be due to current capability, distortion or voltage swing.

With all this in mind, let’s interpret the power data:

  • With both channels driven at 8-ohms, we see 19.7W versus 16.2W. The left side is getting close to the max voltage swing available. The right side falters somewhat, due to higher distortion. Hence it meets the 1% distortion point at modestly lower power.
  • Then, with a 4-ohm load, we get 28.9W versus 16.9W. What has happened is that the right side can’t manage the higher current required, while maintaining low distortion. The left side is now delivering 70% more current than it was with 8-ohms, still maintaining 1% distortion. The right side can only manage 44% more current. However, if we take the distortion limit out of the picture, by just looking for clipping, we find that the right side can provide 22W. Taking the limit out doesn’t help the left side because it maintains low distortion much closer to clipping.
  • power supply capWith just one channel driven, we get a surprise that the channels are about the same at 8-ohms and lower frequency—22.1W versus 21.2W. Apparently, with the additional power supply voltage of single channel operation, both are able to maintain 1% distortion, close to the rails. At 20Hz, it looks like the minimal 2000uF of power supply capacitance (seen at the lower right in the photo above) is having an effect. Power for both channels drops to 15.5W, due to the power supply droop during signal cycles. Since it is voltage-limited, the current limitations of the right channel do not come into play. Increasing power supply capacitance could restore full power at 20Hz, for the left channel.

    However, at 20kHz, the right channel falls apart, yielding only about half of its 1kHz power. This is attributable to the fall off in beta of the old power transistors at high frequency. Their Ft is 0.75MHz, versus 4MHz for the new ones. This means that, if the beta of the old transistor is 30, it begins rolling off at just 25kHz. To deal with second and third harmonic distortion of 20kHz, the loop needs to handle 40kHz and 60kHz. The extra phase shift of the output stage rolloff degrades the feedback, as well as the gain rolloff. The new transistors have 5X the bandwidth. With a 4-ohm load the power limitations are exacerbated. However, as before, if we throw out the 1% distortion limitation, we can get substantially more power out of the right channel, as given in the footnotes.
  • When we move to a 4-ohm load, with one channel driven, the 34.6W versus 19.3W result at 1kHz shows the current limitations of the right channel. At 20Hz though, both channels are hit with the power supply ripple, due to the limited filter cap, managing 16.6W and 8.9W, respectively. The right side is hit with the double whammy of ripple and distortion limitations. Finally, at 20kHz, the left side, with its high beta and high Ft holds up 31.9W, while the right side manages only a paltry 3.6W. To be fair though, if we allow 2% distortion, instead of the 1% limit, the right side can provide 19.4W—a much more face-saving number.
  • To summarize the power results: With the new transistors, the amp can put out about 22W per channel into 8-ohms and 35W per channel into 4-ohms. With both channels driven, it can deliver almost 40W and 60W into 8 and 4-ohms. The specs rated it at 30W and 40W, respectively. With additional power supply capacitance, it could do so pretty much across the audio band.

Discussion of the Other Results

distortion iconDistortion
Looking at lower level distortion (at about half max output voltage), we find the left doing about 2X better at higher frequencies, for the reasons discussed about power. Interestingly though, the left channel is actually slightly worse at 20Hz! Distortion is rising at low frequency due to power supply ripple associated with the small (2000uF) power supply filter capacitance. The amp has limited power supply rejection ratio, so the ripple can induce distortion.

Another factor which can be coming into play is the fact that the bootstrap cap, C303 has a break frequency with R305 at 1.6Hz. For the bootstrapping to be maximally effective, the voltage drop across the cap must be very low. For every factor of two that it is reduced, the effective resistance due to load resistor R306 is increased a factor of two. That increases the gain there by a factor of two, unless it is already limited by the ZL of the output and driver stages (mentioned above). So as the operating frequency gets closer to the break frequency of the bootstrap, distortion can increase. Once distortion is limited by failing bootstrap or power supply ripple, the positive effects of the better output transistors may drop out of the picture. Increasing the power supply cap and the bootstrap cap should help 20Hz distortion.

freq resp iconLine Input Frequency Response and SNR
Due to the resistor tweaks which were added to improve tone control tracking, flatness turned outnoise icon left extremely well, with only a 0.25dB max deviation. Actually the final performance came out even better than the development data, by the simple expedient of finding a better knob position for zero. The 87dB line SNR is also excellent, resulting in virtually silent speakers, even when you are close to them.

riaa iconPhono Preamp Performance
With the mods discussed in the text, RIAA accuracy turned out to be top-notch. At 87dB or better the SNR is phenomenal, considering that the original 2N3391A transistors are being used. Recall from above, that we had a problem in the line stage input, due to current noise interacting with a very high input impedance. As is common, we measured phono noise with the input shorted, which eliminates current noise. I will have to express the caveat that it would have been better to have measured it with a phono cartridge load.

However, since a typical cartridge impedance starts at just 1kohm or so at low frequency and might start rising at a couple hundred Hertz, the troublesome flicker noise we encountered before should be mostly gone at frequencies where the impedance gets very high. Moreover, the phono input transistor is biased at very low current, which reduces input current noise. Nevertheless, it is a concern. If it’s a problem, you could replace Q101 with a 2N5089, as was done for the line stage.

square wave iconStability and Capacitive Loads
The amp drives capacitive loads of 2200pF with almost no overshoot. A 10kHz square wave into 4 or 8ohms looks perfect, with no overshoot or ringing at all. The low frequency stability issue with the power amp has been fully tamed, showing only 1.4dB of peaking at about 7Hz. Even that is no-doubt suppressed by AC coupling in the line stages.

judgementSummary of the Test Results

With the new output and driver transistors, the Cortina 3070 substantially exceeds its power specs, delivering almost 60W total into 4-ohms. With the old transistors, it easily meets the power specs of 30W into 8-ohms and 40W into 4-ohms. However, at 20kHz, the old transistors generate some harmonic distortion at high power. It’s not like anyone can hear a harmonic of 20kHz, though! At half output voltage, distortion is 0.1% for the new transistors and 0.22% for the old ones. This rises for both, to about 0.25% at the frequency extremes, except for 0.4% for the old transistors at 20kHz. Frequency response is within ±0.25dB, 20-20kHz and noise is better than 86dB down, even through the phono preamp! RIAA equalization is within about ±0.4dB, 20-20kHz. Stability is excellent.

The loss in high frequency power for the old transistors should be of no practical consequence since there is little power in music up there. All of the other measurements are exemplary. We conclude that, with the mods which have been presented, the Cortina 3070 does indeed justify Eico’s description as a “Total performance” amplifier; even more so, with the new driver and output transistors.

Listening Tests

Okay, I’m not big on the “wine terms” that some people use to describe the sound of amplifiers but I did a careful listening evaluation of the 3070 and these are my impressions. All tests were done with CD source material. Connecting the Cortina to my 88dB-efficient, 4-ohm, AR-9 speakers, I was really impressed with how quiet it is. Not a trace of hum, even with ears up against the speakers. Hiss is very, very low. [It’s amazing that you can even hear less-than-1mV in a speaker.] Extended listening tests didn’t show an audible difference between the channels. Any differences in power handling capability were lost in the facts that peak passages are short (allowing extra power supply voltage) and high frequency power is low in music.

While I had to set the volume carefully, just under clipping, it was quite loud – much higher than I would normally listen. Admittedly, there wasn’t much margin to spare but I was still worried that I might be disturbing the neighbors, in this neighborhood of 3/4-acre lots. DeguelloPeaks appeared to go to about 18V on the scope. That is the equivalent of a 40WPC amp. The bass from ZZ Top’s Degüello really kicked ass! Hmmm, there might be something to that 70W IHF rating <wink>. After lots of high-power listening with these 4-ohm speakers, the amp got warm but not excessively so. It seems that the chassis-as-heatsink concept works pretty well.dark side of the moon

I mentioned above that I was very impressed with how well the amp did with Pink Floyd’s Dark Side of the Moon and that it really delivered a clean and exciting performance. Bass was very strong and clean. The clarity of the clock shop scene at the beginning of “Time” was stunning.

Overall, the Cortina 3070 makes a superb amp when it is shown a little (well, a lot of) love. biggsPlaying selections from many CDs, ranging from The Tijuana Brass to E. Power Biggs on a pipe organ, I found it really magnificent. During listening, I was startled by one really ultra-low frequency drum note on the hatariHatari! CD, which I hadn’t noticed before. I guess almost all of my listening to that album has been on the vintage system. Those bass reflex speakers are pretty good down to 30Hz (-3dB at 39Hz) but fall sharply below that. Listening on the AR’s and the Cortina really brought it out and it had a big effect! Movie theaters systems which could reproduce low frequency effects would give the audience a special thrill with that. Guess Henry Mancini knew what he was doing!

Secrets of Where the Design Came From

When I restored the Heathkit AA-21 early solid state integrated amp, I was surprised to discover that the power amp had been lifted from the RCA SC-12 transistor manual. Actually, drawing on manufacturer’s application notes was pretty common for the kit makers, as they had limited engineering resources and transistors were new then. Knowing that Eico had even more limited engineering than Heath did, I found myself on a quest to discover where the circuit came from.

It seemed pretty clear that it must be from RCA, having used all RCA devices from their 40000 series. The Cortina line was introduced in 1967. The 15W and 25W circuits in my 1966 SC-12 transistor manual are more primitive, employing transformers. Dr. Lin2I was curious about what the SC-13 1967 edition had, so I ordered a copy from Abebooks, through Amazon. Meanwhile, I found from Jack Ward’s,  that Dr. H.C. Lin (shown at right in 1953) patented the quasi-complementary circuit in 1959, while at RCA, as noted here. [The circuit in the patent shown there is oddly drawn.] A circuit somewhat similar to the Cortina 3070 (but scaled for 6-watts) appears in his contribution to the 1960 book, Selected Semiconductor Circuits Handbook. A note there says that the material was based on a 1956 article by Lin in Electronics magazine. Photo courtesy of the Transistor Museum — used with permission.

When the RCA SC-13 Transistor manual arrived, I eagerly scanned the applications section, looking for the Cortina 3070 power amp but I didn’t find it in the higher power amps. Then I noticed the “High-Quality 10-Watt RCA 13-10 circuit thumbAudio Power Amplifier,” on page-485 (at left). It’s a dead-ringer, except that the supply voltage, driver and power transistors are different! Ah ha! Then I looked back at SC-12: on page-416 was a similar, 10-watt amp, except that the complementary Darlington part is different. So the Cortina 3070 power amp circuit apparently came from the SC-13 circuit, which evolved from the SC-12 circuit, which was based on Lin’s pioneering work on quasi-complementary power amps dating back to 1956! (I feel like James Burke doing the PBS series, Connections :) No doubt that Eico had to use RCA transistors in it, to get free use of the 1959 patent!Mystery-R

Ending with a Mystery...

One component that I have not been able to explain in the circuit of the 3070, is R311 in the power amp (at right). The fact that it is only 4.7-ohms seems to be a mystery. At such a low value, it is difficult to see how it can be significant in the driver circuit. If we compare the 3070 to the original RCA circuit, all of the resistor values in the driver and output stage match, except that one. RCA had it at 4.7K! Could it be that Eico somehow overlooked the “K”, and it has remained undiscovered for all these years? Actually, 4.7K would not work well in the scaled-up circuit as Eico has it. There would be insufficient drive to the output transistor.

I’m not really certain what the designer was after with the resistor, so I can’t pin down a good value for it right now. Perhaps it was intended to limit current in the PNP driver, for sharp, negative transitions, before the output transistor has a chance to respond. Perhaps it is a way to limit loop gain within the complementary Darlington loop and thus contribute to stability. Whatever it is supposed to do, it seems unlikely that 4.7-ohms would do it. However, since it acts a lot like a direct connection, it’s harmless, enough. I will leave this one as a mystery for you, dear reader, to ponder :)

Copyright © 2011 by Stephen H. Lafferty. All rights reserved.



Linda Silk Purseb low