Signal Volt meter, discrete solid state. 2015.
mv-meter-TA-6-front-March-2013.JPG
Picture of front panel, 2015.

This page is about
the Vac meter I designed and built using an old analog
meter and rotary wafer switches re-cycled from old test gear made in
1950s. There are 3 "op-amps" which are built using discrete small size
modern TO92 package bjts and j-fets, plus modern Si diodes, R and C.

The above front panel and the aluminium box behind was previously
used at ANU for a meter to monitor vacuum hardness.

Images below include :-

Fig 1. Resistance divider.
SHEET 4, Block diagram for range switch, 3 amps and 1 meter.
Fig 2. Meter Dial.
SHEET 1. Range switch and Amp 1 details.
SHEET 2. Amp 2 details.
SHEET 3. Amp3 and meter details.
SHEET 5. Power supply.

Why build your own signal voltmeter?
It is very good training for the mind. What good are you if you wish to
make new electronic gear without being able to build a good signal
voltmeter which works as well just as well or better than products for sale costing
hundreds of dollars?

What makes a good signal voltmeter?
Most voltmeters meters on the market are "digital" multimeters, DMM,
and usually hand held units powered by a small 9V battery. They have
numerous ranges and functions and readouts. All these are very easy to use,
especially types with auto range selection, like Fluke, a very good brand.
Advantages :-
1. Light weight and size.
2. Large number of functions, Vdc, Vac, Amps, continuity, peak and hold, diode
and bjt properties, frequency, inductance, capacitance etc, etc, etc.
3. Resolution to 4 significant figures,
4. Able to be used in "floating" mode to measure across 2 live circuit points.
Disadvantages :-
1. Fragile when used with tube circuits; likely to fail easily from high voltages
applied at input, often higher than DMM the unit's maximum ratings.
2. Over time, they slowly lose functions, and become impossible to fix.
3. Cannot read Vac where a lot of Vdc exists.
4. Cannot read Vdc where a lot of Vac exists.
5. Take far too long to settle down to give a measurement.
6. No back lighting to LCD display.
7, Display has many range selections but text size is far too small and decimal
points don't show prominently.
8. Bar meter somewhat inadequate.
9. Cannot read Vac frequencies accurately below 7Hz or above 1kHz
10. Have rather poor input impedance for Vac.
11. 9Vdc batteries go flat too soon, maybe in 1 month with constant use.

The first Fluke DMM I bought in 1993 gave most functions for 20 years.
But the millivolt Vac range died, then resistance measurement died, then
Vac range was erroneous, so it became useless junk for the re-cycle bin.
But it out-lasted about 4 other units I had including Digitech from Jaycar
which was the worst.
The Fluke had most of the advantages and few disadvantages.
My replacement
Fluke has a few more useful functions and is acceptable.

BUT, even the new model 117 Fluke has stepped backwards in some areas.
When using Vac auto range the meter cannot read 0.0Vac, and you
always see 0.022Vac minimum displayed. When using the mV range
for Vac, any presence of Vdc obliterates the measurement of Vac, so
you must use a 0.01uF cap in series with red lead to measure low Vac.
Then readout of low level Vac becomes slow, no better than the awful
Digitechs etc. And you cannot read below 3mVac accurately.
One must get used to a new item where obvious shortcomings are never
described in manuals or data for the item.

To avoid replacement of 9V
battery so often, I use 6 x 1.5 D-cell batteries
soldered together in series to make a single large long life 9Vdc battery.
These are then wrapped with insulation tape between two sheets of thin plywood
to make them well insulated from anything on the bench or in equipment.
You could also explore use of rechargeable batteries.

Be VERY CAREFUL measuring any high voltages!!!!!
It is always dangerous to work on tube circuits even when there is only +/- 200Vdc pk.
All Vac above 100Vrms can be dangerous.
When working with transmitting tubes such as 845, 813, the anode supply Vdc
may exceed +1,000V. These voltages are HIGH and LETHAL. All octal output
tubes such as 6L6GC, EL34 or KT88 can have combined anode peak Vac and
Vdc exceeding the 600V typical max input voltage allowed for the meter.
To minimize meter damage, always Vac at anodes from B+ input at an OPT
to one of the anodes, while monitoring Vac from OPT sec with CRO.
So you always need to know what to expect at primaries. A single 845 making
24Watts into load of 12k0 generates 536Vrms, or +/- 758Vpk.
A PP pentode amp without a load may produce +/-1,500Vpk swings at each
anode and if B+ = +550Vdc, peak swing above 0V can be +2,200Vpk.
You have been warned!


YOU must THINK before making any measurement!!!!
You must know exactly what you are probing, and estimate what could be
the highest voltage likely to be found. To work on tube gear you really need to
have a couple of fixed resistance dividers on circuit boards in plywood cases
and fixed on panels near the work area to allow meter leads to be connected
to low voltage output from divider and probe leads then taken from high Volt
input for divider. You may think a 10:1 reduction CRO probe is both sensible
and handy, but they are no good for HV. Much better insulation is possible
with a fixed resistance divider.

Excessive Vdc or Vac can be instantly destroy a DMM.
All test gear should have input protection preventing damage when Vin exceeds
1,000V.
I once accidentally applied a pulse voltage over 2,000Vpk to a solid state
LabTech CRO. The repairs cost far more than what I'd paid for it, and the
complexity meant I could only keep replacing chips and bjts etc until it worked.
I paid two other techs but they could not get all functions to return return and
sine waves looked distorted. This CRO then developed more faults so it went
out with the rubbish. I repeatedly burned the output bjts in a solid state
Topward function gene. I repaired it several times but it finally fully died in
cloud of smoke and bad smell after allowing its output to contact 240Vac
mains for 2 seconds. My own protection circuit failed to protect against
such accidents.

So I forced myself to learn good workshop habits.

For example, Do not clip meter leads between tube anodes and 0V.
Try to avoid using any leads with alligator clips, especially uninsulated types.
Do not allow any ends of leads to gear to lay unplugged on bench.
If its not being used, unplug the whole lead.

For many, the use of a stand alone cathode follower in a box can be a good
buffer between all signal genies and amp inputs or outputs.
This can be arranged to have many megohms of input Z and high bandwidth.
It is also possible to make a fully floating buffer unit with mains supply also
floating with well separated primary and secondary mains windings.
The cathode follower output may drive a 10k:1k0 transformer with insulation
rating 4,000V, so that Vac readings across two active circuit points at different
Vac and Vdc may be made, while the transformer sec always remains at 0Vdc.
But iron cored transformers have limited Vac range and bandwidth, and add
some distortion to Vac. 

When "bread-boarding" a new circuit use RCA socket for input placed so
stray contact to B+ or high Vac output is highly unlikely. I often make sockets for
2mm or 4mm probe leads using say 6 turns of 1.2mm dia copper wire and
soldered to the 0V rail on board. So risk of the OOPS moment is dramatically
reduced.

Most of the electrical shocks I had were from unguarded mains input wiring.
Although the item is turned off, there are often live active terminals before the
mains switch and which have not been covered with protective shrouding.
Always place covers preventing contact to live mains inputs.

For HV measurements without shock or wrecking meter, here are details
of a resistance divider so that a possible 4,000V can be reduced to a maximum
of 400V.

Fig 1. Resistance divider.
High-voltage-10to1-R-divider.GIF
The above resistance divider can be made using a small plastic box, 9 resistors,
1 capacitor, and 8 recessed 4mm banana plugs, and a couple of tag strips.
The input leads to divider must have good insulation and can be *good quality*
voltmeter probe leads sold as replacements or spares.

There are 2 options of either approximately 10:1 or 100:1 reduction of high Vdc
or Vac. 
The exact reduction ratios will never exist initially because the Rin to any voltmeter
loads the R divider output, and the V ratio increases.

If you use 10:1 outlet of above divider with hand held DMM Rin with Rin = 1M0,
and XCin = 100pF then you have (R7+R8) // 1M0 = 666k / 1M0 = 400k,
so Vout ratio = (6M4 + 400k) / 400k = 16.0:1. It is always possible to make a
R divider give exact V ratio 10.0 : 1.00  or 100.0 : 1.00 with only one of your
DMMs at output by adjusting fixed R values of R7 and R8. Don't try to use pots
for the adjustment. Using just ONE meter only means reliable measurement.
Other types/brands of meter won't give the same reading if their Zin is different.

So your R divider is a PIA to set up properly, but we have avoided the active circuit
or a transformer.
To allow a variety of meter/s and/or CRO to be used from output of divider,
a
cathode follower could be used with its grid direct coupled from output but you
must still adjust R values to compensate for slight drop in Vo because CF gain
with say 6SN7 / 6CG7 never 1.0, but maybe 0.95. A paralleled 6SN7 with idle Idc
of say 8mA will give cathode Rout < 250r, so if following test gear load = 100k,
then V drop is a negligible 0.25%. The dynamic range of the follower may be
limited to +/- 150Vpk, so a buffer cathode follower isn't perfect.

With all reductions of Vdc or Vac input with an R divider, the accuracy is
always challenged and you are lucky to get less than +/- 2% error.

Just remember that where Vin = 3,000V, the V across each of R1-R6 = 450V.
Current in 1M0 = 0.45mA, so Pd = 0.20Watts, so hence R need to be 1W rated,
and metal film, and be able to withstand 450V without mysteriously going open.
Do not use cheap carbon film R.

TWO inputs to R divider.
Top input includes Vdc and Vac but bottom outlet has 0.015uF in series with R1-6
so that the LF pole will be at 1.76Hz.

Many R dividers made for Vac do NOT give a flat response above say
1kHz because C reactance ratio of C between input and Cin to meter is different
to R ratio. To flatten the response to give a high HF pole at say 20kHz, additional
C must be connected across R1-6, R7, and R8. Such C are "frequency
compensation caps" which need to be determined by experiment If your meter
is set for 10:1 ratio and had C-in = 100pF, then expect to have 11pF between
HV input and Vo. 

My Fluke has Rin = 5M0 and Cin < 100pF. Exact Cin is unknown, but one
way of setting divider properties is thus:-
Connect 47pF 630V rated across each of R1 to R6, and assume there is 3pF
across each 1M0. 
This gives you 6M0 // 8.33pF, and XC = 6M0 at 3.2kHz. The C across R7
should have XC = 600k at 3.2kHz so C = 83pF, so if the voltmeter Cin = 30pF
and other C was 5pF then the added C = 83 - 35 = 48pF so add 47pF.
The C across 66k will be 750pF, so add say 680pF and perhaps you'll see a
flat response using a CRO from each divider output up to perhaps 100kHz.
But above 3.2kHz, the divider input Z reduces at 6dB per octave so that at 32kHz,
input Z = XCin = X 8.3pF = 600k ohms. If we assume the signal source being
measured has Rout < 10k, the declining XC at divider input won't change the
Vac reading if you had a wide bandwidth voltmeter.

In practice, to make an R divider with high R values to give a flat response
to say 500kHz IS DIFFICULT, so first use 6 x 47pF each rated for 630V
across each R1-6. Use
say 10Vac square wave input at 1kHz and 10kHz
and m
onitor Vo with CRO. This will show high peaks on wave verticals.
Use AM radio tuning caps across R7 and R8 and adjust for best looking
square wave without overshoot. Variable caps can be replaced by fixed C
with added minor value trim caps.
Of course with Cin to divider = say 8.3pF,
the input Z at 100kHz = 200k, and this may affect the voltage being measured.
However, performance between DC and say 20kHz is usually acceptable
if circuit Z < 10k0.

Never assume the Zin for your voltmeter. Measure this, know what it is!

When the divider is working OK, you should first find that when your Fluke
measures 100Vac/Vdc at HV input, you will read 10.0V and 1.00V at the
two outputs.

There are CRO probes available with 10:1 or 1:1 selectable V ratios.
These are often better made than anything you might make and have
have 10M input Zin so that there is 9M + 1M and the 9M is at the tip of probe
and it has low Cin . CRO probes set for 10:1 can be used after 10:1 divider
for measuring say 3,000V.
There is usually an adjust screw on CRO probe to trim C across 1M0 to allow
variation of C for best square wave with a given CRO connected, ie, give a
good HF response.

It is difficult to measure
low level RF voltages in high impedance tube circuits, eg,
at anode of 6BA6 in AF radio at IF amp with 455kHz. The added C from any probe
slightly de-tunes the IF transformer to alter Vo. Using a follower buffer with j-fet or
tube with Cin < 5pF is wize. Then Vac output is little changed and low follower
output impedance allows viewing signal on a CRO, and without detection happening.
If the CRO is well calibrated, then you can easily see a 10mV RF signal.
However, as F rises, it becomes more difficult to measure RF. But all Vo up
to 200kHz are easily measured in AF circuits.
This website is not meant to fully explain RF phenomena or techniques.
 
Lower R values in divider could be used where circuit impedance at DUT is
less than 10k.
The divider R1-6 could be 6 x 100k in series with R7 = 60k and R8 = 6k6.
Higher HF pole is possible, and DMM with Rin = 5M0 across 66k has little effect
and error < 1.5% without trimming R7 value. Error across 6k6 is negligible.

If measuring Va at EL34 output tubes, the R-divider could be 500k + 56k,
and this gives 10 : 1 voltage reduction. If Vdc was say 500V, then Idc flow from
anode to 0V is only 1mAdc.
Voltage sag at 56k with 5M voltmeter connected reduces expected Vac readings
by < 1% an error which you may/may not ignore. If your voltmeter had Rin = 9Meg,
then sag in measured Vac is negligible.

Having say 3 different voltage dividers each with different R values is not a bad idea.

To measure Vac I often use an oscilloscope, aka cathode ray
oscilloscope, or CRO, because it has a vacuum tube within to display wave forms
without telling lies, so you SEE what you are doing. I have dual trace Hitachi from
about 1983 and a dual trace
Tektronics 465, both nice to use with reliable solid
state to drive cathode ray tube. Bandwidth is DC to about 15MHz. But these
are fragile.

The CRO is not capable of accurate measurement numbers, but for very many
audio measurements we wish to know the F response of circuits so we need to
be able to quickly find the -3dB pole, and see what rate of attenuation or boost
exists. We can view phase shift between input and output with dual trace,
and see the onset of distortion above about 2%, all much faster than listing say
10 Vac levels from a meter.
One might use a PC with LCD screen, or use a hand held device with small LCD,
but the old fashioned CRO with wider BW is better for me. Cheap second hand
CROs are plentiful.
Most DMM have very limited bandwidth useless below 7Hz or above 3kHz.
A CRO set to dc function can measure Vac from DC to MHz.
A good voltmeter has BW = 1.4Hz to 250kHz at least, and there should be an
outlet for connection to a CRO to see what you are measuring.

I built my first bench Vac meter in 1994 with mains PSU in small box with SS
bjt discrete circuit with 6 ranges form 0-10mV to 0 - 1,000V. Bandwidth was not
too bad at -1dB at 2Hz to 200kHz, with some extending to 1MHz, but HF response
above 200kHz was not flat. There was one amp using bjts with gain = 100.
Input Rin to switched input R divider < 500k. The Vac was converted to Vdc
passively to drive the meter. Measurements below 1.0 on the scale of 0.0 to 10.0
were guesswork. I later added a j-fet input buffer to increase the Rin for low
voltages and a more reliable protection network. 

Between 2013 and 2014, I totally rebuilt the meter according to schematics below.
I retained the same old analog meter with its 100mm wide dial face with enough
room for three scales, 0 - 10, 0 - 3.16 and a Db scale, immensely valuable for quick
response checks to about 3 significant figures.

FUTURE IMPROVEMENTS if I ever get time :-
Floating balanced input. This can be done now by using a 1:1 or 10:1 AF transformer at
input with a nominal Z ratio > 20k : 20k, and able to accept 10Vac input without core
saturation. Such a transformer can easily be wound but each P and S winding must have
Lp = 300H+. To get BW from 20Hz to 20kHz means Lp and Cshunt and leakage L must
all be low to avoid a typical response peak between 10kHz to 50kHz. If Fo = 20kHz, then
A Zobel with say 20k plus 470pF at sec may reduce the peak to a fairly flat AF response.
At 1kHz, the Lp Z dominates Zin character and expect Zin = maybe 1M0.
I have a couple of IST, but they do not have sufficient BW, although are useful for measuring
low Z circuit points in amps but must have a series C to floating primary ( ungrounded ) to
keep out Vdc between the two points measured, because at DC, the primary winding
input has only wire resistance and low Z.

I also have a hand held DMM powered by +9Vdc from a mains plug pack PSU from mains.
The LV PT sec is well insulated from PT primary so making floating Vac or Vdc measurements
between any two circuit points is easy without use of batteries.
-------------------------------------------------------------------------------------------------------------------
AC Voltmeter details. 
SHEET4.
general block diagram.
mv-meter-TA-sheet-4-block-diagram-6-March-2013.GIF
SHEET 4 shows a 12 position wafer switch is used with 3 amplifiers to power
an analog meter. The content of SHEETS 1+2+3 are within dashed lines......

Case = 425mm wide x 135mm high x 250mm deep, 1.6mm Aluminium, except
front plate = 3mm thick. Heatsink for PSU regulators is Al plate between
PSU and amps. The whole of SHEET 1 and SHEET 2 and the rear of the meter
itself is encased in internal steel box to give some magnetic shielding and ensure
low noise with amps having input impedance above 20Meg even when using
the 0.0 - 1.0mVac range. 

My meter dial can be copied into your PC image program and changed in size
to perhaps suit your meter. Most analog meters have swing of more than 90
degrees as mine shows. The original size is actually larger than shown.
Fig 2. Meter Dial.
mv-meter-dial-printable-6-March-2013.GIF 
I made a template using white cardboard and pencil to plot the 3 scales
using my Fluke to verify voltage measurements. The scale is substantially
linear, because the Vdc used to drive the meter is derived from the GNFB
network of a meter amp.

For changing Vac ranges, I used an old double rotary wafer switch with 12
terminals for 12 positions. These old switches commonly used terminal No12
which was also a pole which points to itself at position 12. So I could get only
11 Vac ranges from 0 - 1mV to 0 - 100V.
The 0 - 316V is possible by having a non switched 4mm banana socket for only
316V when the 100V Vac range is selected.

The meter is calibrated so that a pure 400Hz sine wave at 10.0Vrms gives
the same reading as my Fluke digital meter.
Other readings above and below 10V are consistent with a good DMM.

I did consider rigging up a 2.50Vdc reference diode so peak Vac of a Vac
source could be adjusted for 2.5peak, ie, 1.767Vrms, and the meter amp
gain adjusted so needle sits on 1.767 Vrms on 0 - 10V scale.
But the Fluke seems to be accurate enough, and its better to have two
good meters which give the same Vac reading, even if they are slightly
in error by up to +/- 2%. 

To allow floating Vac measurement just like a hand held device I would
need to use a transformer with floating primary winding as mentioned above.

SHEET 1. Switch details and Amp1.
mv-meter-TA-sheet-1-input-sw-x10amp-6-March-2013.GIF
For the low Vac ranges 1mV to 100mV.
The 5 low level Vac inputs are fed from input RCA or banana socket through 10k
bypassed with 15nF. This R&C prevents excessive input Idc flow in limiting diodes
at Amp 1 and Amp inputs. The input blocking caps charge up slowly.
The 10k is thus a protection measure. Once all coupling caps charge up within
the unit there is very little delay waiting for Vac reading to settle.

The input is fed to terminals 1a to 5a and then to input of Amp1 which has a j-fet
2SK369 with bootstrapped bias R21, so input Z = 20Meg with some shunt
capacitance of mainly Cg-d of 15pF with maybe 5pF of other stray C.
For ranges 1-5, Zin = 20Meg bypassed with 20pF.  
The source follower connection reduces the Cg-s from 75pF to negligible amount.
So, any input signal will see 20Meg at 50Hz but the 20pF reduces Zin at 6dB / octave
above 400Hz. The Zin is about 1Meg by 8kHz. For most amplifier measurements
the Vac being measured has source resistance below 50k, thus allowing HF -3dB
pole at 159kHz, assuming the source is not already shunted by any additional
capacitance.

Amp 1 increases signal 10.0 times, and feeds its low Z output to R divider
R3 to R7.

Amp 2
selects signals via pole to points 1b to 5b. At each Vac range the Amp 2
input is a maximum of 10.0mV max for the 5 ranges.
Amp 2 increases signal 10 times to provide a maximum of 100mV to power Amp 3.
Amp 2 input also has a bootstrapped source follower with Zin about 15Meg
bypassed with 20pF.
 
Amp 3
converts the Vac to Vdc linearly to work the meter for full swing.

Output from Amp 2 of up to 100mV can be viewed on oscilloscope.

For all input signals above 100mV, Amp1 is not used and input is directed to each
separate R divider for each of 6 Vac ranges. Each of these R dividers has
Rin = 3.06Meg, with much less R as the series R.

For all Vac ranges from 0.316Vac to 100Vac the Zin = 3.1Meg bypassed
with about 4pF.

There are many capacitors which need critical adjustment. C3 and C23 are
small ceramic trim-caps set to minimize any oscillations above 1MHz.
These are very likely if there is poor layout, inputs close to outputs, or if electro
rail caps are not bypassed with plastic caps and if leads are not all kept extremely
short. I've used a number of 220r "gate stoppers" to prevent spurious HF above 1MHz.

Caps C9,12,14,16,18,20 are made with a 12mm length of 1mm copper wire soldered
to switch lug with small piece of 0.5mm insulated telephone hook up wire soldered
to other switch lug, and then wound around 1mm wire until response from a
signal gene showed best HF extension without peaks or troughs, and a good looking
40kHz square wave.

Caps C11,13,17,19 were chosen after each of the others was set for about 3 turns
of wrap, ie, about 4pF. Once the response looked nearly flat and square waves had
little peaks or rounded corners the wire wrap caps were adjusted for best flat response
and best square wave.

I was able to get all ranges from 1mV to 100V to give bandwidth giving -1dB at 1.4Hz
to -1dB at over 250kHz with less than +/- 0.2dB change along each band.
The source signal comes from the low impedance output from a completely re-built
1980 BWD function generator with F output from 0.1Hz to 2MHz.

SHEET 2. Amp gain x 10.
mv-meter-TA-sheet-2-x10amp-2-monitor-6-March-2013.gif
Amp 2 is explained well within text on the schematic.
You may find this amp could make a splendid line level audio preamp.

SHEET 3. Meter Amp 3.
  mv-meter-TA-sheet-3-meter-amp-3-6-March-2013.GIF
This meter amp uses 4 germanium diodes in a bridge rectifier to create a
small Idc flow to power the meter.
This diode bridge and meter is within the NFB network with R11, R14, VR1.
The total value of these three resistors = 73r. If you make this circuit, the R values
will need to be different to suit the meter used.

The GNFB action makes the conversion of Vac to Vdc linear. The FB virtually
eliminates the the non linear forward voltage turn on transfer function of diodes.

Calibration of the meter is by applying 100mV to Amp3 input, and turning shaft
of VR1, 300r, so that the meter has a full swing.
VR1 is a 25mm pot mounted on the board and its 6mm dia metal shaft to a
plastic shaft with screw slot and protruding through front panel where its not
likely to be disturbed. Once set, the calibration has shown no sign to drift.
The NFB eliminates all non linearity including temperature drift.
Replacement of the meter would require the NFB resistance network values
to be revised.

C3&R5a, C8, C9
are needed to prevent HF oscillation above 5MHz.

Probably, the smart arses among you will laugh at my primitive circuit.
But you may find ordinary op-amps would be easier, until you realize most
do not have the ability for such wide bandwidth.
It would be nice to have HF -1dB pole at 5MHz or higher.
But once you get to RF, measurements are difficult unless the circuit
impedances are much lower. This Vac meter is meant for the audio tech
needing to know about signals between DC and say 500kHz. 

SHEET 5. Power Supply.
mv-meter-TA-sheet-5-PSU-6-March-2013.GIF
There is nothing unusual about this generic PSU. The R3,4,5,6 47r plus bypass
caps seem to fully suppress any stray coupling at any F between the 3 amplifiers
which would cause serious stability problems.

Back to Education and DIY directory.

Back to Index page.