As mentioned in my article on the Kenwood KX-880HX, the early and mid-80s saw a small number of 2-head decks from various manufacturers that were built around the '1C-DD' single-capstan version of the Sankyo direct-drive mechanism renowned for being used in the Nakamichi CR-7/5/4 ('2C-DD'). These decks were a curiosity in that they married a relatively high-end transport to a 2-head configuration, and not the usual 3 heads.
Truth be told, 2-head decks can be excellent, and e.g. Technics made quite a few fine direct-drive machines. The 2-head Nakamichis from the BX, CR, Cassette Deck, and DR-series also were rather good. These latter were based on the Sankyo single-capstan belt-drive mechanism ('1C-BD'). This raises the question if ever there was a Sankyo direct-drive machine that could compete with these Naks. Finding the answer is hard, because of the relative scarcity of these decks. But I was lucky: early in 2021 the rarest of them all, bar perhaps the little Kyocera, found its way to my door: the Rotel RD-870.
I picked it up from an elderly gentleman. Unable to wait I dropped it in my car's trunk and opened the cassette well. That head looked ... funny. Once at home I took a closer look and yes, the head was not the original sendust type, but a non-descript small-form permalloy one. A nasty surprise!
There is not much information on the 870 available, but the few pictures on the internet strongly suggest that it uses the same Canon head as the Nakamichi 2-headers. While I have a spare CR-2 transport in store, I deemed it unwise to sacrifice it. At any rate, its head has a bit of wear, whereas I wanted to give the Rotel the best possible chance to prove its mettle. I was lucky enough to source one of the last new old stock Nak-labelled heads on the planet.
If we take the Sankyo mechanisms of the late CR-era Nakamichis as a template (i.e. CR-1/2/3/4, 1988), then the one in the Rotel diverges somewhat. The reel drive is via gears instead of the older wear-prone rubber idler, making the transport 'late' in the evolution. Yet, the reel brakes are the rubber-tipped metal claws from the earliest versions. Both the gears and the brakes of my deck are as per the service manual, not later add-ons or modifications. Also curious is that there is no external device for back tension: uncontrolled tension is provided by a leaf spring at the bottom of the supply reel axle.
I believe that Onkyo was the original and earliest customer of the Sankyo family of transports: TA-2070 double-capstan direct-drive in 1981, TA-2055 single-capstan direct-drive in 1982. That same year the latter was also found in the Kenwood KX-880, Kyocera D-601 and Sansui D-370/570 (these with solenoid control instead of a cam motor), and the single-capstan belt-drive variant debuted in the Nakamichi BX-1. (I don't think anyone else ever used that same mechanism: Onkyo's belt-drive Sankyo-based machines used some plastic parts instead of the all-metal Nak versions.) The Rotel dates from much later again, 1986. Incidentally, the Sankyo brushless direct-drive capstan motor is also used in the TEAC V-1RX and V-900X!
Being direct drive, with gears, and without tension belt, the Rotel RD-870 is as rubber-less as can be and should be dependable as well as a doddle to service. And yet, I found wow&flutter inconsistent. It could be low, but it could just as well be high, especially towards the end of the tape: 0.075% and worse. Speed stability could be fine, or it could significantly slow down towards the end, accompanied with chaotically rising w&f.
Weeks of trying with various test tapes, disassembling, relubricating, and reassembling everything was to no avail. I changed R374 to a 10k trimmer (preset to 6k8), making reel torque adjustable as per Nakamichi practice, then setting it at 40 g.cm. I tried four different pinch rollers (two new, two used), on three different assemblies. In the end I replaced the back tension leaf spring with the coil spring arrangement from my spare CR-2: curiously, the Rotel had all the right holes already predrilled. This seemed to help somewhat. (The graphs were made before and after the modifications, using my TEAC MTT-211NA speed tape and WFGUI software.)
The deck also was incapable of playing a mirror tape: it skews at the roller, regardless of which roller or assembly, head present or head removed. It does play a homebrew inspection tape properly, though. (This is a cassette with most of the lower shell removed, but with but with all tape guides still in place.) With nothing left to replace and nothing to adjust I am at a loss as to the root cause of this.
Mounting the brand-new head was easy, just taking care to solder quickly enough to avoid damage to the coils. Replay level could easily be set (ANT L04 Dolby level tape). I checked frequency response using TEAC MTT-356 and Hanspeter Roth alignment cassettes (the latter are remarkably good and very affordable). After azimuth adjustment both were consistent with each other. The RD-870's response was essentially extended, but with the treble shelving down from 2kHz on.
In my view there are two acceptable ways of playing such reference tapes. Ideally the response is ruler-flat. Nakamichi took the view that the response must be rising above 10kHz with frequency; they had their reasons. This policy has always been controversial ('IEC 1981 or not'), but it has the undeniable advantage that it fights treble loss. And we all know that cassette loses treble whenever it gets a chance: azimuth error, Dolby mistracking, tape and head wear, .... Some brands took the opposite view, shelving down or rolling off replay treble. This was probably done to gain an unfair advantage in signal to noise ratio, or to provide brand lock-in.
Anyway: I bought the Rotel out of curiosity, but also to serve as replay deck for difficult, old, suspect, or worn tapes. This demands a good and neutral response. So even if I would have loved to restore the RD-870 to its original state, I was now forced to compensate that treble curve.
Similar to the Kenwood the Rotel has a 'bath tub' construction: the circuit board cannot be accessed without major disassembly. Luckily this is much easier than with the Kenwood. It suffices to remove the PCB's many screws, cut one ground wire, and fold the PCB out at the deck's front. Awkward, but workable. To protect the cosmetics I removed the front panel and knobs, and stuck styrofoam blocks to the front, aiding the deck in standing up during soldering jobs. Even so this arrangement is not conducive to repeated trial and error work, so I had to do all circuit mods in one go, and only then reassemble the deck and test the result.
In a first approach I worked on the main time constant of the replay amplifier, changing R105 and R106. (Incidentally, the service manual is full of errors. Beware!) Increasing these resistors over their initial 15k value brings up the midrange and treble. I then effectively added a Play Trim-like function by relegating the duties of R105/106 to a newly-made board with a 10k stereo pot and four resistors, as well as jumpers to disable the PT circuit when not wanted. To control the PT pot one would have to open the deck, but that is not an impediment to my intended use. (Curiously, there is a commercial precedent of a deck with internal-only Play Trim. I forgot which one, possibly Kenwood or Luxman ...)
This contraption worked well enough, but in the end it did not satisfy because, by the time Play Trim was turned up enough to compensate for any high treble loss, the midrange became unbearably bright.
A solution was found in adding an inverse treble shelf to the playback amplifier. This was done by putting 180R + 100nF in parallel each to R103 and R104. Thus I lifted the treble by ~1dB at 4kHz and 2.5dB at 10kHz. After this the improvised Play Trim subjectively worked better.
The first graph shows the response for both channels with Play Trim set to the minimum position. The highest treble is still rolled off, but the overall response is now quite flat over the bulk of the frequency range. (Ignore the treble loss in the left channel: this run was made with the last set of tones on the Roth tape, close to end of tape, where some skew is possible.) The second graph shows the right channel for three positions of the Play Trim knob.
At this time the deck no longer could be considered a pure RD-870. But the mods can easily be undone, by plugging in two jumpers and by cutting two links from the top of the PCB. No disassembly required!
Back to the internal Play Trim. When facing the deck the replay amplifier sits right after the front-mounted Rec Balance and Bias Fine controls. These controls are two potmeters mounted on their own little PCB attached to the front panel. With the preamp that nearby it would be very feasible to redesign this little PCB, now to carry the Fine Bias pot and the add-on Play Trim circuit using the hole and knob of the Rec Balance control (which has little use in most circumstances anyway). This would be a very neat solution. Maybe in the future, should I decide to keep the deck ...
One curious aspect of the various Nakamichi decks that use the same sendust head is their lowish Maximum Output Level. I have seen 3-4dB over Dolby level for type I and II, and often less for type IV (where MOL normally would exceed the other types). I have to assume that Nakamichi deliberately underbiased in order to get a better treble response (without HXPro), with the customer expected to employ Dolby C in combination with low recording levels. Nakamichi's philosophy centered around classical music (much like Quad), so this seems to make sense. What we don't know is if that Canon head itself demanded such low levels (of bias and of music), or if in reality it could be pushed harder. This was another reason for me to try the Rotel, to get an alternative view on the same head.
But as said before, my pulling up the replay treble response moved the deck away from its initial design parameters, so we will never know for sure.
The rising playback treble has to be balanced with reducing recorded treble. This can be done by tuning (or rather: redesigning) the record equaliser for each tape type, or by increasing bias. The former is the correct approach, at least when standing bias has been selected judiciously, but it would be a lot more work than the latter.
Increasing bias initially did a fine job here, but when I came to metal type IV the treble could not be flattened: it kept rising, indicating significant underbiasing. The graph below is for 1992 Sony Metal XR, with the internal bias and the front-mounted Bias Fine both maxed-out. There is a 3 to 4dB lift at 20kHz!
Investigating I found two things. Like many other HX Pro-equipped decks the RD-870 uses the NEC uPC1297 IC for bias control. This chip takes as input the raw bias signal, the two music signals as fed to the recording head, and two static DC signals signifying the desired static bias level. The latter DC signals are controlled by the tape selector switches, the front-mounted Bias Fine knob, and the 3 x 2 internal bias trim resistors.
The internal type I and II trimmers correctly influenced the mean bias over their full range. But the type IV trimmers did not effect any change in the upper 60% of their travel. It appears that the Rotel's metal bias control sits above the saturation level of the HX Pro chip's voltage controlled amplifier. Moreover, where the service manual prescribes a +14.4V supply for the replay amp and the bias control, I found only 12.6V. Changing R302 and R303 brought this up to 14V and this then produced a little bit more bias signal from the uPC1297, but really not much. The HX Pro VCA saturating for metal type seems to be a design error. With no obvious way for boosting bias I had to give up on type IV.
The resulting record/play curves for type I and II are really quite flat and extended, considering the changes done to the replay side.
At 20dB below Dolby level (actually -19dB) type I has its -3dB point at 18kHz, and at Dolby level at 9.7kHz. For type II this is 20kHz and 10kHz, similar to a Nakamichi CR-2 and much better than the Kenwood KX-880HX and NAD 6100 Monitor I once had.
The 400Hz MOL figures are +5.2dB and +5.5dB, respectively, indicating excellent behaviour from the head, better than the Nakamichis. The deck's no-tape noise floor is -60.7dB(A). That is OK, but not fantastic, the result polluted by supply hum harmonics. The real-world noise with an old BASF single-layer chrome was -60.1dB(A). Thus with a decent ferro-cobalt tape and no Dolby a dynamic range around 62-63dB would be possible, really not bad.
As said before metal remains underbiased. The -19dB curve extends to 22kHz, the 0-dB curve to a ludicrously high 17.6kHz. Due to the low bias MOL is a poor +2.5dB. I don't know for sure, but I have a feeling that even in its native, non-modified state the 870's metal MOL would be low. Just as with the BX and CR Nakamichis.
There is only one set of internal record level trimmers, so compromises are necessary. I settled for aligning to 1994 Maxell UR, a tape with a sensitivity, see https://audiochrome.blogspot.com/2020/12/index-to-cassette-tape-measurements.html, close to the average of 80s-90s basic and premium ferrics, as well as the two IEC I references.
1987 TDK SA (itself close to the post-1987 IEC II reference) then gave +1dB offset, and 1992 Sony Metal XR +0.3dB. Both are acceptable values and still useable with Dolby NR. (If you wonder why the frequency plots are only starting at 200Hz, and not at 20Hz or so: this makes synchronising the https://www.audiotester.de/ measurement software easier.)
The Rotel is a fairly large deck. At 433 x 115 x 300 mm it dwarfs everything here except the CR-7. Its overall aspect is rather blunt and brutish, and not very elegant. It has only few features, and thus few controls, big controls at that. Being more spartan than even a Nak CR-1 there is no output volume control, no headphones output. A pity: headphones are handy for quickly checking the sound during repairs. Now I had to connect it to an amplifier whenever I wanted to listen. (There is no music system on my work table, but I have a TEAC AI-101DA 'bible amp' for grabs between the technical books on a nearby shelf. Of course I still have to grab it and wire it up to the DUT when needed.)
The door is a flat piece of hard glass-like plastic, screwed into the cassette well frame with two hex bolts. So you need a hex key whenever taking off the door for cleaning! The transport switches off when there is no tape in the well, but this can be fixed by wrapping sticky tape around one contact pole of the tape sensor switch.
The light bulb behind the beauty plate was broken. I replaced it with a green LED, its current set to 18mA by means of a 750R resistor in the positive supply line. This was not a success: the LED sits below a diffusor meant to spread the light into the tape well. The LED does not emit much light upwards, so the overall effect is barely visible in daylight, and that is even without a tape loaded. If you attempt such a mod consider a LED of higher power, and possibly a white LED.
The counter is outright daft. During winding or play it counts reel revolutions. During recording it counts in real time! Both counting modes are not correlated: zero in play mode is not the same as zero in recording mode. That is inconvenient, because doing tape alignment on a 2-head deck requires synchronisation of the recorded signal to the computer, calling for an accurate return-to-zero. With the Rotel this is provided by the absurd sequence {push zero, enter record-pause, push zero}. Even then the accuracy of the Rotel's counter is such that when returned to zero it is 5-10 seconds off the true initial position. (That may have to do with my operating the transport mechanism on top of the deck, exposed to incandescent light. For sure during the restoration the transport and counter behaved differently depending on which lights were on in the room!)
The LED peak meters allow for even less adjustment than the Kenwood KX-880HX: none at all! The preset level for meter-zero appears to coincide more or less with Dolby level (215 nWb/m DIN = 200 nWb/m ANSI). There is a slight imbalance in my deck's meters, with left favoured over right. Luckily their granularity is a coarse 2dB around zero, so this is not too conspicuous with real music signals. The meters make recording on the selected tapes easy: just peaking at +4dB resulting in fine results both on type I and type II. I did not try type IV.
Conclusion
The Rotel RD-870 has its quirks, but even so it is a fine deck. Once the drooping replay response has been taken care of it is a worthy competitor for the various two-head Nakamichis, even though it is less versatile and a bit harder to use and maintain.
