Tube amplifier El Cheapo
Page 2: Modding
Last updated: Monday March 8, 2010
Let's start with the original circuit and look where we can gain .. (click on the drawing to enlarge)
However, this schematic of Suppo audio is not complete nor correct. I'n not claiming to have the complete overview myself, but in any case I've drawn up the following shcematic which is absolutely more accurate than that of Suppo Audio. Most important changes: The 6 capacitors of 220 uF in the power supply are in actual only 4 caps of 150 uF (The light red ones are not present on the PCB). The original schematic shows a cathode resistor of 100 Ohms, whereas in real-life these are two resistors of 270 Ohms in parallel (mind, the board is designed for 3 resistors of 390 Ohms in parallel yielding the same resulting value of 130 Ohms).
The following is a list of possible mods:
First thing to improve with this amp is the safety issues. I thought that, since it hadbeen discussed on the diyaudio forum, the electrical safety had been enhanced in the running series of amplifiers. However, when opening up my GV3 version, it was clear that this had not been fixed. Also, the community had complained about the small power switch that was of inferior quality. It was a DualPole switch, but Suppo has soldered both halves of the switch together in order to have more reliable operation.
The cabinet has a IEC connector accepting modern power cords, however it is the simplest version without a fuse holder in it. The power entry socket 
has been "glued" to the chassis, so there was some room for improvement. Please click on the first picture to see the small switch and the original power inlet.
As I soon found out, the GV3 version of the amp that I had differed internally from the pictures found on the web. What I did not know was that the new square transformer, unlike the toroid used in versions GV1 and GV2, forced the two output transformers inside to be relocaed moe to the sides of the box. As a result, there was very little room to position a new net entry or upgrade the power switch.
After looking it from several different angles I decided to position the new power switch on the top plate behind the large black power transformer. This was definately not my favourite spot, but unless I decided to put a power switch on the front plate it was the only logical place (left).
I removed the old IEC socket and replaced it with one with an internal fuse holder. As you can see on the pictures, there is room enough to do so, but it;s not much. I insoluated the soldering joints for the phase and zero connections with red shrink tubing. I did also connect the eart connection directly to the chassis using 1.5mm installation wire. In the original situation, the earth was connected to one of the "min"connections of the power capacitors and from there through the holes in the PCB the safety earth was connected. I do not think this is at all safe, the safety earth should be connected to the cabinet with a short wire (and star washer etc.) directly upon entering the cabinet through the IEC power inlet.
Also shown on the picture is that I found a new use for the small round hole in the top plate where the original switch was located. As the new swich is out of sight behind the power transformer, I will include a small LED so that we can see whether the amp is in operation (should we not see the orange glow of the heaters).
The simple solution for the LED is to use one of the heater supplies (there are 2) and signal that the heaters are on. A slightly more complicated approach is to measure the voltage on the cathode resistor and use that to switch the LED. Find below the circuit to do that. We need to rectify the AC supply and need to connect to the top of one of the two cathode resistors of 270 Ohms.
Hmm, After changing the bias to a CCS circuit the circuit above does not apply any longer as there is no cathode resistor.
I've bought a 21-step attenuator wih SMD resistors. I did have a ALPS 50k volume pot in stock, but it might be just too large to fit in place. The original pot has known issues wrt channel balance and is source of cracks when turning the knob.
On eBay they sell a very small 21-step attenuator with SMD resistors. I bought a 50kOhms log type.
Today I installed the attenuator in the anplifier Since Suppoaudio glued the original pot and the connecting wires with lots of hot glue, there was only one thing I could do in order to keep the amp tidy: Remove the original pot and its wires and reinstall the attenuator and connect it with Prefer Microphone wire. The prefer wire is of medium quality type, it has PTFE (teflon) isolation and is reasonably thin and easy to handle.
It looks like the wire used to connect the input connectors to the volume pot is already of decent quality, but I can replace with good quality (blue) Prefer cable. On the web there are several people who replace this wiring with silver cable, sometimer from Siltech such as Menno van der Veen does in his UL40-s2 modifications.
I've used Prefer shielded microphone cable which was obviously of better quality than the standard wire that it replaced together with the original volume pot.
The original 6N1P tubes are Chinese versions. I do have a few good Russian 6N1P-EV tubes that I want to try out.
I'd like to see whether the 6N14P or 6P43P tube is better quality than the stock 6P15P power tubes. Reading the datasheets, and looking at the circuit, the amp is indeed built as if the power tubes are EL-84 (or 6N14P) tubes, and looking at the name of the amp, and the 6P14P markings on the PCB it sure looks like the 6P15P tubes used in my amp is used for economic reasons (hopefully for sound quality also). The -8V grid bias is conform what other sources on the net use for a pretty standard EL84 PP amp. It looks like the 6P15P tubes is compatible, but there are not so many sources for the datasheets of this tube.
And, there is one thing to take into account: in the 6N14P and EL-84 versions the 2nd screen is internally connected to pin-3. For these tubes, pin-s 1, 6 and 8 are in principle not used. However, as some sources have found out, the fact that the datasheets tells us not to use them doesn not mean that there are not EL-84 variations on the market that have these pin's connected internally to something ...
For the 6P15P tube the situation is slightly different: Here pin 1 and/or 6 are used to connect screen 2. On the PCB of the amp, pin 1 is connected to pin 3. So this is the 6P15P way to connect screen 2 via pin 1 to pin 3.
Therefore, as pin 1 IS necessary to be connected to pin 3 on the PCB, for any tube to be compatible with this amp pin 1 must be absolutely NOT connected to any other pin (other than pin 3 or to screen 2).
The following tubes are "more or less" compatible with the original 6N15P
Only problem for this upgrade is the limited space we have to get these capacitors installed. I do have styroflex caps available, and with the change of the grid resistors from 200K to 1M (or 470K) the lower frequencies won't suffer.
Of course we can always replace the existing coupling caps with exotic and expensive types such as Auricaps. I do not know whether it will for this amp improve the sound that much but for sure it won't hurt either. However the price tage is too much and not in relation to the rest of the amp. The improvement much be quite dramatic, for me to go that path.
There are some calculations that help to determine the correct values for the coupling caps. For example, the lowest frequencies that pass the filter are determined by: 1 / (R C )
Ripple rejection in the B+ lines can be improved by adding an additional filter. According to people on diyaudio.com who measured the B+, it is 284Vdc for the output tubes and 245 Vdc for the drivers. This means that 350 or 400V capacitors are the best choice for any capacitor upgrade in the amp. At this moment, the power transformer outputs about 230Vac which is then rectified and filtered only once with 4 capacitors of 150uF. Only the driver tubes have a 2nd RCR filter installed which will help in reducing humm in the input circuit. Playing around with PSU designer (duncanamps.com) it shows that adding a RC filter consisting of a 470uF capacitor and a 220Ohms resistor will remove ripples in the B+ lines (now 3Volts) to 50mV. Only the B+ voltage drops about 23V which may not be acceptable.
Well, we can decrease the value of the resistor to 100 Ohms which will already greatly help, but maybe I'll be looking for another solution.
So what are the alternatives?
A CHC filter will not fit, as I do not have room for a choke in the case. But we can use a solid state alternative to the choke such as the gyrator. In fact, it seems that several people have already built something like this. And as I'm considering upgrading/modding my old UL40-s2 amp as well I've been looking at the e-choke form Menno van der Veen and got the impression he's using a variation on a gyrator as well. Most designs use zeners for voltage reference, but this means that the gyrator is dependent on a fixed and set voltage. The mentiond e-choke and some other designs are not but reference the input voltage instead which makes more sense in my opinion.
Links:
The heater supply, which is at this moment 6.3V AC, seems to be doing OK for this amp so I'm not going to change it for the moment..
Don't know whether faster diodes will actually improve the sound. If I would choose to add an additional filter in the supply, then I need to get the diodes off the board anyway.
I do not think that I'l go for the product of Guido Tent here (as much as I like his solutions) since with it's price tage of 179 Euro it is more expensive than my entire ampshipped from China to my door. The Tent solution probably is of very good quality and it manages 4 tubes, where at this moment the bias is set for each pair of output tubes combined. There are comparable solutions available from Matchmat (Matthijs) and others but they do not differ in pice that much that I'm prepared to cough up the money. After all, this project must remain a true El-Cheapo amplifier.
But, autobias would be my preferred solution as through the grid voltage the tube will bias itself to a preset cathode current.
So maybe if we won't use autobias we maybe can use a Constant Current Source (CC
S). Not the same as auto-bias, as it will not bias the output tubes the same way.
I've seen a couple of these from others already, the most simple one being a voltage regulator with a resistor in series with the output. The idea here is that the voltage regulator will always try to keep the voltage on the adjust pin at around 1.25 Volts higher than the output. So by putting an additional resistor in series between the output and the adjust pin we can compute the value of the resistor knowing that the voltage difference between the two will be 1.25 Volts.
The LM317 is the most simple / and perhaps elegant / solution. According to the National datasheet, I_out = V_ref/R_adj. So in this case we can compute R_adj as follows: R_adj = 1.25 / 0.060 (2 tubes of 30mA) = 22 Ohms. Should I want to use a slightly higher bias for the tubes, then a resistor of 20 Ohms would do just fine (around 35mA per tube).
Well, and if we would like to make the design perfect, we could balance the load of the 2 tubes with a small pot OR bias each of the 2 tubes separately (which is at the moment not done).
So, I've removed the two cathode resistors of 270 Ohms R11, R20 and the two cathode capacitors C4, C16 (there are 2 caps per channel) and replaced them with a CCS consisting of a LM317 voltage regulator and two 47 Ohms resistors in parallel. On the picture below you can see how I soldered the CCS on the PCB. I removed both cathode resistors and capacitors and used two of the holes to solder the adjust and input pins of the LM317. The middle pin, which is the output pin, is bent up and connected to two 47 Ohms resistors in parallel. Of-course I checked before soldering that the cooler and the LM317 together would fit under the PCB without touching the cabinet. And also I checked whether the construction would bent easily or not, well it turned out that even with just two feet attached to the PCB the LM317 will stand off just fine.
I've listened to this new setup and I must say that it works remarkable well. Of ourse we're adding a semiconductor device and putting it in the cathode circuit of our audio amp, but I think in that position it's not going to add to the sound character too much. And the original setup, with a resistor and a capacitor to ground in parallel is also not a very good solution either since a capacitor will add to the sound as well and it's behaviour for lower frequencies is not ideal also (and for larger value we might have to bypass with a smaller value).
Of course the even better solution would be to give every tube it's own CCS regulator. This will make sure that influence of individual tube aging will be further minimized. But since both cathodes are not connected anymore, we have to bypass the bias circuits with a sufficiently large capacitor.
I'm very satisfied with the setup so far. Only next time when I'm opening the case I'll put some shrink sleeve over the two resistors just in case they will bend at some moment in time and touch other parts on the PCB
I've been reading several sources on the web that use the LM317 as a CCS in their tube projects. However, regarding the use of a bypass capacitor from Cathode to ground I've not found a conclusive answer whether as to use such a cap (electrically I can do without, but sonically I don't know). Keeping things simple, I decided for the moment not to use a bypass cap.
Alhough UltraLineair operaton is not possible with the current output trannies, triode mode is. It will decrease the output power of the amp with 3dB, so I need to find out whether this is what I want. Well first of all, we'll have to inspect the board to see whether there is an easy way to connect the 2nd grid to the plate. Well, it turns out that we have to cut around pin 9 of the output tubes the traces on the PCB to the B+ power lines. For 3 tubes this is easy to see and can be done with a small Dremel tool (make the cut at least 1 mm wide). For the 4th tube the trace goes under the tube socket and has to be cut at the other side of the socket. ALternatively one can remove the socket from the PCB but this is a he** of a job since the soldering power needed to get things connected to large "islands" on the PCB is far too much to be practical.
Some others on diyaudio.com have done this already so I guess that unless I have a completely different board this will be possible without too much work. If so, the question remains whether the sound quality will improve at all and whether the power remaining in the amp is sufficient to drive my Tannoy speakers. The two bookshelf Mission M70 speakers I use at the moment are only 87dB in sensitivity which means that this is a less than ideal match with the amp.
Anyway, I took the risk and have cut the connection between pin 9 and the B+ on all 4 output tubes. After that I soldered on the bottom side of the PCB 4 resistors of 100 Ohms between pin 7 and pin 9 on all 4 output sockets. And as far as sound quality is concerned I can report that it doesn't hurt to do this mod. Less distortions at higher volume levels.
The amp becomes hot, very hot, after some time (hours) of continuous operation. Although this is not an immediate issue as I've never met a power amp that I can touch after one hour of working, it's better to control the temperature.
One easy way is to drill ventilation holes in the amp, preferrably both in the top plate as well in the bottom. Drilling holes in the top plate will yield the best results, however I'm not looking forward in drilling holes in a 3mm plate which is also spray painted. Well, since looks and WAF is not an objective for this amp I might as well drill those holes.
Below you find the circuit as it is after me making my modifications (in blue). As you saw at the top of the page, the official Suppo schematic is not entirely correct and contains some errors (in red). Most obvious: The power supply capacitors of 220uF are in reality only 150uF and instead of 6 there are only 4 of these caps present.
Apart from making modifications I've been listening to the amp quite a few times in my study. But I'v tested in the living room as well. Using the Marantz DV12-S2 top SACD player which has 2 front outputs it was quite good possible to connect the SACD player to two sets of amplifiers.
As you can see I listened to Sonny Rollins in Vienne. BTW, only the small speakers connected as the large Sonus Fabers were connected to the Marantz PM14ki amp.
| << Page 1: Vanilla |
^^ Back to DIY Page ^^
|