Despite finding at least 1000 online retailers selling this 'kit', there's not much technical info written about it, so I'm going to let everyone know what they're in for should they be interested in buying one. It's based on an early kit produced by the German electronic kit company, Kosmos, called the "Radiomann". Outside Europe it's given a more Anglo sounding name, "Radio Ace".
I bought mine off eBay from Online Science Mall, back in February 2008. This was one of the cheapest retailers at US$79.95. Postage to Australia was unbelievably fast too; I had the radio in about a week. Many thanks also for this company sending its products overseas...always pleased to promote establishments that do this.
It is rather surprising that so many retailers continue to sell this receiver at over $100, so it pays to shop around.

Construction
Although promoted heavily as a kit, it is anything but. The radio is already assembled and all that has to be done to put it into operation is to insert 8xAA cells, screw the aerial coil under the three terminals, plug in the headphones and connect aerial and earth.
The chassis is made of veneered MDF and stained in a walnut colour. All parts are new. Despite the modernity of the construction, it is very attractive and really does look the part.
Under the chassis, the wiring is point to point. No PCB or tagstrips are used. Wires are soldered to the ends of components with no other support.
Two lengths of clear cellulose tape are an attempt to keep the wiring attached to the underside of the chassis. Four sets of double AA cell holder are bolted to the back of the chassis.
Plastic binding posts are used to make aerial and earth connections, and also to connect the aerial coil. Two pre wound coils are supplied with the kit; one spider web type wound with Litz wire on a blank PCB former for medium wave reception, and the other a conventional air cored enamel copper wire solenoid for short waves wound on a plastic tube.
Modern low impedance headphones with a 3.5mm stereo plug are supplied and are fed via what is obviously a mains transformer to match the phones to the 12AU7.
The tuning capacitor is an air spaced metal unit, obviously meant for a superhet receiver. The oscillator section is not used.
Quarter watt resistors and ceramic capacitors are used for the circuit.
A miniature 50K switch pot of the type used on transistor equipment is used to switch the 12V supply and adjust regeneration.
The valve is a new Chinese 12AU7 and the socket is porcelain.
All in all the parts are perfectly good quality.

Simple design using point to point wiring. Even though there is no metal chassis, hand capacitance turned out to be insignificant.

The Design
The Radio Ace is a very conventional regenerative receiver using a 12AU7 twin triode. One triode functions as a regenerative grid leak detector, resistance coupled to the second triode which is an audio stage feeding low impedance headphones via a transformer. This transformer has, according to the circuit diagram, a ratio of 20:1.  It is clearly a mains transformer and would be 220 to 12V given the German design. Provided there's sufficient turns per volt and the DC flowing through the primary isn't too high this scheme works very well. A 100V audio line transformer would be the other preferred option in keeping with using standard modern parts.
The unique aspect of design is that the B+ is only 12V; the same supply as used by the valve heater. Supply is from 8x AA cells. With the heater drawing 150mA, it is wise to use alkaline cells for anything but short periods of use. At 12V the B+ current draw is insignificant. For use of NiCad or NiMh cells, and extra double AA battery holder should be put in circuit given these cells provide 1.2V instead of the usual 1.5V from carbon and alkaline cells. A DC input socket would have been a worthwile feature to allow extended operation from a mains supply (or in my case, the 12V home lighting plant). The manual claims the set works even when the battery has dropped to 9V. This turned out to be true.
The aerial is coupled to either the entire coil or to the tapping by various values of capacitor. The choice of connection depends mainly on aerial length. In areas with strong signals, spider web coils will pick up sufficient signal not to require an an external aerial.
This set uses plate voltage control for adjusting the regeneration.
Feedback is from the detector cathode to a tapping on the aerial coil. This eliminates the requirement for a separate feedback winding.
No volume control is provided; instead it is expected that the regeneration is backed off if volume needs to be reduced. However, doing this has the disadvantage that selectivity suffers. In practice, with a set like this it is seldom that the volume does need to be reduced.

Book Review
It is obvious that the writings in the manual are a translation from German. Some of the terminology would be strange to those not familiar with German radio. Not once was the receiver referred to as "regenerative" or "Reinartz", let alone "TRF". Instead it was called an "Audion", which in English is the name of Lee De Forest's triode.
Nevertheless, it's a nicely set out manual and an interesting read. The 30 "experiments" are fairly limited and most of them would be applicable to any other radio. They involve things as trying different aerials etc, touching the coil to damp it, and observing how the regeneration control works. Only towards the end do the experiments become slightly more technical with such things as creating positive feedback through the audio stages using the resistance of one's fingers to couple to output to the input. There is also some mention of winding your own coils and adding bandspread, although not much practical detail is gone into.
I'm not trying to be critical here; just alerting would be buyers of the Radio Ace that is more of a built up radio to be used as is. It is not, as one might think from the advertising, a kit you install the parts and solder together, or even a kit where the same components are connected in different ways to create different circuits.

Trying it out
The radio was unfortunately an immediate disappointment. I've had many years of experience with regenerative receivers and this one was among the worst. Although regeneration was very smooth and operated as it should, the gain and volume were very poor. In fact, the volume was no better than that from a crystal set. Despite their obvious cheapness, the headphones are of reasonable quality and I verified this by using them with another receiver. However, they are not the most sensitive of this type of headphone I've tried.
The aerial connections seemed to be limited; "Ant 1" and "Ant 2" were next to useless and "Ant 3" was the only connection that was usable. This was with long and short aerials; the long being my outdoor wire aerial of about 30m length, and the short aerial being the few metres of wire supplied.
I have operated vacuum tubes at low voltage before and knew that this wasn't the cause of poor performance. There should be much more volume into headphones from a 12AU7 operating with 12V on the plates.
The poor performance with the aerial connections was actually expected; capacitive coupling into an aerial coil is poor practice. Not only does the aerial load down the coil, but sensitivity at the low end of the band is reduced compared to the top end. A better way is to have a separate primary winding.
The regeneration operated very smoothly, which I did expect. I am pleased this set uses plate voltage control for adjusting the regeneration. This or adjusting the screen grid voltage where tetrodes or pentodes are used is the smoothest method. Controlling regeneration by means of a variable capacitor is not only makes for critical adjustment, but has bad backlash and worse, detunes the receiver as the control is adjusted. The other popular means of adjusting regeneration by shunting the feedback winding with a variable resistor is even worse. While the receiver isn't detuned by this method the backlash is considerable and adjustment extremely critical.
Clearly, some modifications would have to be done to make it a practical receiver with good performance and not be relegated to being an attractive static display.

Original circuit of the Radio Ace. I thought some of the component values were a bit strange even before I tried out the set. My suspicions were correct.

Redesigning the Ace
The choice of component values had me curious and it seemed the designer must have had some knowledge of valve regenerative receivers, or managed to find circuits to copy, but some of the values chosen would appear more suited to a transistor circuit than a valve one.
First thing to do was fix up the grid leak; 100K is a rather low value and there was no way as much detected audio can appear across it compared to a much higher value. Normal values of grid leak are from around 500K to 2M. I tried 1M which improved things markedly; but 2.2M was even better. The value of grid capacitor is a bit high; something like 100 or 250pF is usual. I left the 330pF in situ as the difference doesn't make it worth changing.
Next was the grid leak for the audio stage. Again 100K is low and is not used for this application except in certain high gain output valves which are prone to grid emission...something the 12AU7 is not going to experience running at 12V! Increasing this resistor to a more appropriate value of 1M gave further improvement to volume.
Incidentally, in case you are wondering about lack of negative bias for the output stage, the 12AU7 triode cannot draw significant current at 12V and anything more than a small amount of bias would impair the operation of this stage.
Next improvement was for the detector. 22K as a plate resistor is more appropriate to solid state circuitry. Low values mean low gain. It was necessary to increase this resistor at least to around 47K; up to 220K if possible. Doing so with only 12V B+ might however reduce the detector plate current to the point where the detector won't oscillate. I had thought of schemes involving 9V batteries to jack up the voltage, but as it turned out this wasn't necessary. With a 100K plate load the detector oscillated just as well as before, even with the aerial coil heavily loaded. The difference now was we had much more volume and gain. It was what this type of receiver should have.
Final mod was the aerial coupling. While I would prefer a separate primary winding for the aerial coil, to add one would be impractical. An extra terminal would be needed on the aerial coil just for starters. So, I was left with trying to make the best of the capacitive coupling. First thing was to remove the connection to the coil tapping as connecting aerials here was useless. Worse, connecting through the 12pF capacitor was even more of a waste of time. This is an absurdly low value to use and is more suited to VHF receivers. The coil tapping is at a very low impedance and a 12pF capacitor has a very high reactance at 1000 kilocycles, meaning virtually no signal gets through. I settled on a direct connection to "Ant 1" to allow for very short aerials, via the 33pF to "Ant 2" for medium length and via the 12pF in series with the the 33pF to the "Ant 3" connection when long aerials are used. The problem with capacitive coupling to aerial coils is that tuning, selectivity, and regeneration are very dependent on aerial characteristics. It's very tight coupling and certain aerials can actually stop the detector being able to oscillate; not only that this may be evident at only certain parts of the band. Further, even when regeneration can be optimally set, selectivity can still suffer because of excessive signal input due to the tight coupling. The other problem is aerial capacitance is effectively in parallel with the aerial coil and thus affects tuning range. A long aerial can prevent stations at the top end of the band being received. Yet another bad aspect is that gain across the band varies; high at the high end where the coupling capacitor reactance is low, to poor at the low end when the reactance is higher and less signal is coupled. This method of aerial coupling strikes me as being for people who are too lazy to put an extra winding on the aerial coil. If one insists on this method, it's best to use a variable capacitor, but that means another control to adjust each time a new station is tuned in. One single winding with the aerial capacitively coupled does seem very popular with simple receiver designs and I wonder if those who use this method get frustrated at the poor performance.

Redesigned circuit. Modifications were: - R3 to 1M, R2 to 100K, R1 to 2.2M, Ant.1 direct to C1/R1/C4 junction, Ant.2 to Ant.1 via 33pF, and Ant.3 to Ant.2 via 12pF.

The New Improved Ace
The receiver is now a pleasure to use. Volume is more than adequate now and the DX performance is what it should be; for example receiving 2XL from Cooma in Sydney at good volume.
The lack of volume control is not really an impediment as volume is not overpoweringly loud. In Sydney, it is only necessary to earth the receiver to hear the local stations at good volume; an aerial is not needed. However, 2m of wire connected to "Ant 1" will give even greater volume. In the Blue Mountains, the full 30m of outdoor aerial connected to "Ant 3" gives good results. Unfortunately, the limitation of not being able to receive stations at the top end of the band becomes evident when doing this. It would be necessary to reduce the number of turns on the aerial coil to compensate. However, some coverage at the low end of the band would be lost doing this. The seriousness of the problem is not sufficient to warrant modifying the aerial coil.
To the listener it certainly isn't obvious the receiver is running off 12V. It performs much like any other one valve set and the good volume doesn't give it away.
Incidentally, there is enough room under the chassis for a 12V 2.2Ah SLA battery which would be a more economical way of operating the receiver away from the mains.
One convenient aspect of this receiver is having the binding post aerial coil connections; it thus becomes a very handy test bed for trying out other coils without having to desolder anything. Another feature is that provided one completes the grid circuit, by removing the aerial coil, the set can be used as a low power headphone amplifier. Connect the audio input between "Ground" and "Ant. 1" (in the modified receiver, or to the aerial coil terminal that connects to the tuning capacitor in the unmodified set). If the audio source does not have a DC path to earth, simply connect a resistor across it. The value isn't critical; something around 1M will do. Using the set like this isn't anything new; it was often done to add gramophone pickups to early radios.

Other Valves
Given there's quite a few other twin triodes with the same pin connections, I decided to try some other types to see how well they worked in this regenerative circuit at low voltage. First off, the other well known 12.6V heater types; 12AT7 and 12AX7. Not useable at all. Some faint sound was heard but that was it; no regeneration was possible.
Then to a couple of 6.3V types, with the heater temporarily powered from an external supply. 6ES8 is a frame grid VHF amplifier used in TV tuners. As I suspected, due to it's low plate voltage (90V) and very high gain (12.5ma/V) it worked well; certainly just as well as the 12AU7. 6CG7 was the other valve I tried. It's actually the 9 pin replacement for 6SN7; to be specific the 6SN7GTA. It was intended to be used in TV line oscillator circuits, but like other valves found widespread applications elsewhere. Dsepite its similarity with 12AU7, not only did it work very well, I suspect even slightly better than the 12AU7. I think the 6CG7 works so well with 12V plate voltage simply because of it's hotter cathode (6.3V@600ma) which is twice the power of the 12AU7's heaters. I do wonder also if the larger cathode and plate area also helps. No doubt 6SN7 should work the same, being the equivalent. Presumably 12BH7 would be another suitable candidate. However, if operating off a limited battery supply, the obvious choice is to stick with the 12AU7 as all the other types draw more heater current.

12V High Tension for valves
Seeing as I haven't got around to it yet and this is the most appropriate place to discuss it, I'll give a quick mention about 12V operation of valves. This is nothing new. Operating valve plates and screens off 12, 9 and even 6V is not new. Not only was this done with one valve sets just like this one to avoid a separate, expensive, and large B battery, but from the late 1950's to the mid 1960's the technique was used in hybrid car radios. Valves were used for the RF, Converter, IF, detection, and  audio amplifier in the normal way, but running off 12V B+. Because valves cannot provide high power output at 12V, a transistorised stage was used to drive the speaker. There is sometimes the assumption that the valves used in hybrid car radios were all of the space charge type. Space charge mode is used with so called "dual grid" valves like the type 49 and later 12K5. Here, positive voltage is put on the first grid (equivalent to the control grid) to force an increased electron flow. The second grid (equivalent to the screen grid) is used as the control grid. Some pentodes, in particular the 6C6, have been connected to work in space charge mode in simple one or two valve sets running off 6V B+. Apparently, when so used, the heater voltage becomes rather critical and needs to be less than 6V.
Apart from the valve driving the output transistor in some car radios (e.g. 12K5,12AL8,12DL8, 12DS7,12DU7,12DV8), the other valves are in fact not space charge valves. Not only does the circuit diagram prove the point by showing them connected as per normal 250V valves, so does the internal construction. Generally, the space charge valves provide an audio power of 20-40mW which is sufficient to drive a power transistor operating in single ended class A.  However, conventionally constructed valves are also used as drivers; 12J8 and 6ET6 are examples. A few Astor circuits even show a 6BA6. Where a conventionally constructed (i.e. non space charge) valve is used and cannot drive the output transistor on its own, a second transistor is used for driving purposes.
The plate current is minimal when conventionally constructed valves are used on 12V.  What needs to pointed out here is there's a big difference in voltage gain and power gain.
Plate voltage is not so important with voltage gain in the front end of a radio receiver, and in fact reducing the B+ does not cause it to drop off as much as might be thought. On the other hand, power gain is drastically reduced, and using valves like 6AQ5 as output valves with 12V plate and screen supply results in only milliwatts of output. Good for headphone use or even a speaker in a quiet room, but totally useless for a car radio.
The 12V "hybrid" or "car radio valves" (especially those in the front end) are essentially the same as their mains counterparts but either selected versions, or versions made to much higher specifications. Radio & Hobbies for April 1959 discusses the design of hybrid car radios, and gives examples by saying 12BL6 is constructed similar to 12BA6, and 12AD6 is similar to 12BE6. Essentially, the grid characteristics are made uniform for low voltage operation. It's possible to use selected 'mains' valves as replacements, or alternatively adjust the bias to suit if necessary.
To prove my theory, I got out a mains operated radio with the very typical 6AN7(ECH80) and 6N8(EBF80) converter and IF/detector valves.
The audio was a 6DX8(ECL84) triode pentode. I took the 6X4 rectifier out, connected a 12V DC supply into the set's HT line, bridged all the screen and decoupling resistors, and reduced the 6AN7 oscillator feed resistor to half. The heaters were powered off the transformer as per usual. I wasn't really surprised at the very comfortable headphone volume level and the good sensitivity of the set. Sensitivity appeared to be much like any other valve superhet. The set really started to liven up at 15V B+. The limiting factor was the local oscillator being shunt fed by a resistor. Had I changed the local oscillator circuit, the set would have functioned with less than 12V.
Futhermore, once when servicing an HMV hybrid car radio with a broken 12BL6, I did actually subsitute a 12AU6 and it did work. To prove the point, I have constructed a perfectly normal superhet receiver using '250V' type valves, but operating on 12V. See link below.

• Superhet using normal valves with 12V B+

More Information:
There is some interesting and useful info here.
Use the Google translator to read in English here. The site examines using other valves and aspects of the design. It appears that the author is actually the designer of this and other Kosmos kits.