I’m interested in the comments above because while many make sense, it really isn’t the make that is important to range, it IS the power output, but balanced with the receive sensitivity of the receiver, coupled to the available filtering. The actual power output is important, of course, but in many cases, the ‘best’ antenna in use in an in-band system will be the receive one that as said, is usually the higher one, because there’s no point having a good transmit system if it can’t hear anything. Power can be useful because very often not all of it gets to the antenna. Motorola are well known for having pretty decent receive filtering and a clean RF output. This helps reduce the number one in-band repeater problem - receiver desensitisation. Even with good filtering, there will be problems when the receiver is presented with another strong RF signal just a few Mhz (or even KHz away). It swamps the receiver front end and makes it act as if slightly deaf. The amount varies from radio to radio. Motorola is pretty good, but there again so are Icom and Kenwood because they also use the two radio system in their cheaper repeaters. On my own repeater, with the transmit frequency 6.5MHz above the receive, listening to a weak signal on the input becomes a problem when the transmitter fires up. It drops the receiver sensitivity and the weak signal becomes noisy or simply absent. In practice, the repeater fires up when the weak signal appears, but then as soon as the transmitter is up, it drops out again.
You can experiment with aerial systems, and the most common has a vertical for receive and the transmit one below. If you use a folded dipole on a side arm, you can spin it around the mast until the small null from the support mast faces the direction of least interest. It makes no difference to the vertical above. If the receive antenna is a side mounted dipole too, then it too will have a slight null, but almost certainly you will want this to be also in the direction of least interest, the same as the lower TX one. If you squeeze the last dB out of the desensitisation issue, sadly, you also get a small drop in gain in the most useful direction? I’ve never found it makes a practical difference, and in my dual dipole systems I install the dipoles in the same plane above, and use vertical separation to help the desense, but then often have more RF needed to get the receive and TX distance about the same.
With 5MHz or more, I rather like the small compact duplexers for UHF, and use a common aerial with gain, which usually offsets the loss from the duplexer with a bit to spare, and the duplexer solves the desense issue with it’s sharp filters on the TX and RX frequencies. In practice, a duplexer, even with it’s loss gives me slightly extra range than two antennas.
While doing some testing on a system last year, I used a dirt cheap baofeng as a transmitter and RF wise I found no difference between it and the real Kenwood repeater. The receiver sensitivity was also about the same. However, it did suffer from desense more than the Kenwood - I suspect it has very little filtering at all.
When it comes to the actual repeater connections, the ideal circumstance is that the audio coming in is matched in level to what the transmitter needs. The only real circuitry needed is the components needed to switch the transmitter on. Cheap Chinese ones use vox - quite coarse vox, as in silence, or not silence. So any clicks, hiss or a voice switch it on. That’s pennies worth of components and not clever at all. In better (read more reliable) systems all that is needed is a point in the receiver that changes voltage when the squelch opens. I’ve been modifying radios to do this since the 80s, and it’s not hard. LED receive lights are a simple source, but a look at the circuit diagram always shows somewhere you can take a feed - usually via a semiconductor so there is no current demanded - the connection to the circuit being a high impedance one so it doesn’t add any appreciable load to the circuit being tapped (a bit like when people add discriminator outputs to a receiver). Once you have a point where 0V turns into 5, 12 or whatever V when the squelch opens, you can use solid state switching or even a small relay to key up the transmitter. Some radios have an accessory socket, and one pin on these is usually there for squelch open duties.
The big thing to frustrate these kinds of repeaters is duty cycle. Typical mobile and handheld radios are designed to do 15% receive, 5% receive and the remaining 80% as idle. Keying these up for long periods makes them hot and they can easily overheat. Some mobiles can manage up to 50% transmit, and have big heat sinks to remove the heat, often with a fan. These are the best for repeaters of this kind. 'Real" repeaters will have a 100% duty cycle and usually comprehensive output monitoring that will drop the RF if the VSWR changes, as in maybe the antenna or feeder having a fault.
Back in the 80s I modified a fleet of Icom VHF and UHF radios to do crosswind repeater duties for emergency use. I connected the squelch pickup point to the metal belt clip, and the repeater box had two bits of PCB where the radios slotted in - which picked up the 5V line. One dual plug into each one and the repeater simply worked.
The cheap ebay boxes and professional rack mount repeaters all suffer from desense issues, the switching and logic have no impact on this phenomenon.
If you get desense, and it’s easy to test by listening to weak stations when you turn the TX on, then distance between aerials and filters will help the most. Separating the aerials by more than 5m is usually a cure with powers of 10W and below.
My repeater here is designed for low power hand held radios. A 4.5dB white stick on receive, and a 3dB J pole on transmit, with 5W - and NO filter, with the two aerials at the same height but 2m apart horizontally. If I go up to 25W output, the filter is required and this reduces the range back down to the same as the 5W tx range outwards. It seems balanced enough that it’s sensitive enough to give the users a good chance of being heard, without mobiles with better aerials being able to access too far out for licensing reasons.