Poor soundstage with SET 300B

[jeepy]

Hello All,
I hope that someone will be kind enough to help me since I am disappointed with my just completed 2 mono blocks SET 300B amplifiers. Based on a simple schematic: 5842 Raytheon + Mûndorf 0.22 cap + 300B Svetlana + Partridge Xformer (all tubes = matched pairs). All voltages and bias are ok for both channels (180V/22mA) and (410V/70mA) automatic bias. I also used good quality componants: Caddock, Cerafine... Everything is ok except
the sound stage! By exemple, with some Cds that I know very well I can't really locate the singer wich is exactly in the middle at a couple of meters when I use a 'basic' solid state integrated amplifier. The problem is the same with my P610 Diatone and my Harbeth HL-P3. I you have any idea, please let me know.
kind Regards
Jeepy, Switzerland.

[Yvesm]

No LSP phase error . . . of course ?

[jeepy]

Bonjour!
No LSP phase error. It's even a bit worst if change phase on one side.

[SHiFTY]

Do you have a link to the schematic? Might need some tweaking or, dare I say it, some negative feedback added.

[kevinkr]

The quality of the cathode bypass capacitors are critical in an auto-bias circuit. What are you using in these locations?

Your description sounds more like a lack of resolution. Careful listening may reveal that a lot of information you have been hearing with the other amplifiers is just not being reproduced.

I've designed and built a lot of 300B SE amplifiers and they generally image pretty well in my experience - no inherent reason with either of the tube types you have chosen to have this problem.

Can you post the schematic or provide a link to it?

Edit: Can you post some pictures of your amplifiers showing internal construction details such as grounding, component placement and anything you think might be relevant.

[andyjevans]

I'm not sure that you can talk about "soundstage" without talking about speakers. What are they, and have you tried moving them around a bit?

Andy

[jeepy]

Bypass caps are Cerafine 100mF/100V for 5842 and 220mF/100V for 300B.
Bandwith is ok, signal is clean but I noticed that the amps are cliping for 1VRMS sine wave for almost of the frequencies between 20Hz and 20Khz on input wich means 5.3 V RMS (about 15V peaks on the scope) on output with a 7.5 Ohms resistor load. If I am right, I'm just a hobbyst not an engineer, this means that the max power is U2/R = 3.74 W where is amp is supposed to deliver 7 to 8 W!?!
I put the schematic on the web for you. Many thanks.

300b - image / photo - Image perso - Hiboox

[kevinkr]

Originally Posted by jeepy

Bypass caps are Cerafine 100mF/100V for 5842 and 220mF/100V for 300B.
Bandwith is ok, signal is clean but I noticed that the amps are cliping for 1VRMS sine wave for almost of the frequencies between 20Hz and 20Khz on input wich means 5.3 V RMS (about 15V peaks on the scope) on output with a 7.5 Ohms resistor load. If I am right, I'm just a hobbyst not an engineer, this means that the max power is U2/R = 3.74 W where is amp is supposed to deliver 7 to 8 W!?!
I put the schematic on the web for you. Many thanks.

300b - image / photo - Image perso - Hiboox

I looked quickly at the schematic, basically I would recommend taking an oscilloscope and measuring the signal voltages in each stage just prior to the onset of clipping to determine whether the design intent is being met. Are all voltages as shown in the schematic?

You should be able to determine where the problem lies. Incidentally 1Vrms is insufficient based on the available gain to drive the amplifier to full output which would require ~140Vpp on the grid of the 300B, 1Vrms given the basic design would give you about 45Vp (90Vpp) at the grid of the 300B. Something like 1.4Vrms is required and this may exceed the available cathode bias if it is less than the expected value. (Resulting in very poor linearity)

I suspect the bias on the first stage is more than a bit off - some tweaking might be required to get the voltages just right. I've found that high transconductance triodes and pentodes do not always bias that predictably due to the fact that minute changes in effective grid bias can cause very big changes in operating point due to the steep slope of their characteristic curves. IMO conventional cathode bias does not work well with these tubes at all. (I have gone to LED or fixed bias and tweak circuit constants over a small quantity of bogey tubes to get consistent performance.)

Were it me once I understood the problem I would replace the RC based cathode bias with a ~ 2.5V - 3V led (maybe a green led?) and use a CCS in the plate circuit of the 5842 to set the desired operating conditions. Based on my experience this would allow you to swing enough on the plate of the 5842 to reach full output.

Related issue: I have not yet found a large value electrolytic that sounds (or measures) ok with just a couple of volts of dc on it. I have not tried the very low voltage Black Gate which might be ok - or not.

I would start with just a LED for biasing the 5842 and then tweak the supply dropping resistors (not plate load resistor) to get the plate voltage in line, but a CCS or even a plate choke (Helps linearity with large signal swings and can swing above the quiescent supply voltage) are worthwhile experiments. Note that even working optimally there is not a lot of margin for variations in operating point in that first stage. Another tube with more gain that is worth considering is the D3A - I like it even more for these sorts of tasks than the 5842 which is really more suitable for direct driving types that require less than 100Vpp for full output. (An IT with a mild step up changes the picture considerably.)

Note that output power will be slightly reduced if you are using an OPT with a 3K primary, but you would still be able to achieve at least 7Wrms out with a properly working driver stage.

One other comment is I do not see a grid stopper resistor on the grid of the 5842, this is critically important because without it the tube can oscillate at frequencies well beyond the range of most scopes used for audio work. (VHF typically sometimes > 100MHz) A 221 - 1K resistor mounted right at the socket in series with the grid should be used. I have found that most resistors (even metal films) work just fine in this application despite the sensible recommendation for a carbon composition resistor in this location. (CC are non-inductive into the VHF region, whereas the others generally aren't - IME in most cases any resistor except a wirewound is better than no resistor.)

[jrenkin]

You might look at the Tubelab SE for ideas on implementing a 5842 in a 300b circuit. George has done a lot of work for you.

[dsavitsk]

Originally Posted by kevinkr

IOne other comment is I do not see a grid stopper resistor on the grid of the 5842, this is critically important because without it the tube can oscillate at frequencies well beyond the range of most scopes used for audio work.

This surely explains the bad sound. Stoppers should go on all 4 grids. Stoppers should also probably go on the plate and cathode.